A method and system for intelligent horizontal correction of X-band radar antennas

Abstract

This application discloses an intelligent horizontal calibration method and system for X-band radar antennas, belonging to the field of data analysis technology. This application proposes a radar antenna array calibration scheme based on a multi-dimensional state attribute vector set and machine learning algorithms. By deeply analyzing sensor measurement state data, it can accurately identify the X-band performance characteristics of the antenna array and provide specific calibration decision guidance for each performance characteristic. This approach enables intelligent and automated calibration of radar antenna arrays, significantly improving calibration accuracy and efficiency, and providing a more reliable guarantee for the widespread application of radar systems.

Description

Technical Field

[0001] This application belongs to the field of data analysis technology, specifically relating to an intelligent horizontal correction method and system for X-band radar antennas. Background Technology

[0002] In the field of radar technology, accurate and efficient calibration of antenna arrays is crucial to ensuring stable system performance. However, traditional calibration methods often rely on manual intervention and periodic maintenance, which is not only time-consuming and labor-intensive but also susceptible to errors introduced by human factors. Especially in complex electromagnetic environments and variable weather conditions, the performance of radar antennas can be gradually degraded due to various factors, leading to reduced accuracy and reliability of surveillance. Therefore, there is an urgent need for a technical solution that can automatically and intelligently calibrate radar antenna arrays to improve calibration efficiency and accuracy while reducing operation and maintenance costs. Summary of the Invention

[0003] This application provides an intelligent horizontal correction method and system for X-band radar antennas, which can solve or partially solve the technical problems involved in the background art.

[0004] This application provides an intelligent horizontal correction method for an X-band radar antenna, applied to an AI correction system. The method includes:

[0005] Acquire the sensing measurement state data of the target radar antenna array, and determine the multidimensional state attribute vector set of the sensing measurement state data;

[0006] The target performance feature identification result is determined based on the multidimensional state attribute vector set, wherein the target performance feature identification result includes several first antenna performance feature units, and the feature weight of each first antenna performance feature unit indicates the confidence that the first antenna performance feature unit is an X-band performance feature.

[0007] The target calibration feature recognition result is determined based on the multidimensional state attribute vector set, wherein the target calibration feature recognition result includes several second antenna performance feature units, each second antenna performance feature unit corresponds to a first antenna performance feature unit, and the feature weight of each second antenna performance feature unit indicates the calibration decision feature matched by the first antenna performance feature unit corresponding to that second antenna performance feature unit.

[0008] Based on the target performance feature identification results and the target calibration feature identification results, determine the X-band performance features and the expected horizontal correction features;

[0009] Based on the X-band performance characteristics and the desired horizontal correction characteristics, the distribution information and deviation information of the data to be adjusted are determined in the sensing measurement state data, and the target radar antenna array is calibrated based on the distribution information and the deviation information.

[0010] In one implementation, determining the target performance feature identification result based on the multidimensional state attribute vector set includes:

[0011] A feature embedding operation is performed on the multidimensional state attribute vector set to determine the multidimensional state attribute vector set with completed feature embedding, wherein the multidimensional state attribute vector set with completed feature embedding includes several third antenna performance feature units.

[0012] For each third antenna performance feature unit in the multidimensional state attribute vector set with completed feature embedding, the target pointing angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the pointing dimension is determined, the target phase angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the phase dimension is determined, and the feature weight of the first antenna performance feature unit in the target performance feature recognition result is determined based on the target pointing angle feature variable and the target phase angle feature variable.

[0013] In one implementation, determining the target pointing angle feature variable in the pointing dimension of the multidimensional state attribute vector set that has completed feature embedding includes:

[0014] Based on the pointing dimension, the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first pointing feature label of the multidimensional state attribute vector set are traversed, and the maximum value among the traversed feature weights is determined as the first target pointing angle feature variable. And / or based on the pointing dimension, the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second pointing feature label of the multidimensional state attribute vector set are traversed, and the maximum value among the traversed feature weights is determined as the second target pointing angle feature variable. Determining the target phase angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the phase dimension includes:

[0015] Based on the phase dimension, the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first phase feature label of the multidimensional state attribute vector set are traversed, and the maximum value among the traversed feature weights is determined as the first target phase angle feature variable. And / or based on the phase dimension, the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second phase feature label of the multidimensional state attribute vector set are traversed, and the maximum value among the traversed feature weights is determined as the second target phase angle feature variable. The determination of the feature weights of the first antenna performance feature unit in the target performance feature identification result based on the target pointing angle feature variable and the target phase angle feature variable includes:

[0016] The first target pointing angle feature variable and / or the second target pointing angle feature variable are weighted together with the first target phase angle feature variable and / or the second target phase angle feature variable to determine the feature weight of the first antenna performance feature unit corresponding to the third antenna performance feature unit in the target performance feature identification result.

[0017] In one implementation, the target performance feature identification result includes at least one performance attention index, wherein the feature weight of the first antenna performance feature unit in each performance attention index indicates the confidence level that the first antenna performance feature unit is the X-band performance feature of the target of the corresponding performance type.

[0018] In one implementation, determining the target verification feature recognition result based on the multidimensional state attribute vector set includes:

[0019] The multidimensional state attribute vector set is subjected to feature embedding operation to obtain the target calibration feature recognition result. For each performance type of X-band performance feature, the target calibration feature recognition result includes at least one performance attention index. The feature weight of the second antenna performance feature unit of each performance attention index indicates that the first antenna performance feature unit corresponding to the second antenna performance feature unit serves as the performance correction reference under the set correction rule matched by the X-band performance feature of that performance type.

[0020] In one implementation, determining the X-band performance characteristics and the desired horizontal correction characteristics based on the target performance characteristic identification results and the target calibration characteristic identification results includes:

[0021] Based on the feature weight of each first antenna performance feature unit in the target performance feature identification result, the discrimination score of the first antenna performance feature unit belonging to the X-band performance feature is determined;

[0022] The first antenna performance feature element with a discrimination score greater than the set discrimination score is identified as the X-band performance feature; and based on the distribution characteristics of the X-band performance feature in the target performance feature identification result, the horizontal correction expectation feature corresponding to the X-band performance feature is determined in the target calibration feature identification result.

[0023] In one implementation, determining the distribution information and deviation information of the data to be adjusted in the sensing measurement state data based on the X-band performance characteristics and the desired horizontal correction characteristics includes:

[0024] Determine the feature association description between the multidimensional state attribute vector set and the sensing measurement state data;

[0025] Based on the X-band performance characteristics, the expected horizontal correction characteristics, and the feature correlation description, the distribution information and initial noise information of the data to be adjusted in the sensing measurement state data are determined;

[0026] Based on the multidimensional state attribute vector set, a noise perturbation for the initial noise information is determined;

[0027] Based on the noise disturbance and the initial noise information, the deviation information of the data to be adjusted is determined in the sensing measurement state data.

[0028] In one implementation, the X-band performance characteristics and the horizontal correction expectation characteristics include X-band performance characteristics for the data to be adjusted for at least one time-series label and horizontal correction expectation characteristics for the data to be adjusted for that performance type.

[0029] In one implementation, the sensing measurement state data is radar antenna array state data collected by a built-in measurement device. Before determining the multidimensional state attribute vector set of the sensing measurement state data, the method further includes: adjusting according to the original feature values ​​of each antenna performance feature element in the sensing measurement state data.

[0030] In one implementation, determining the multidimensional state attribute vector set of the sensing measurement state data includes:

[0031] The sensing measurement state data is processed by at least one feature embedding branch to obtain sensing measurement state data with completed feature embedding.

[0032] The sensor measurement state data with completed feature embedding is subjected to at least one feature pooling branch and at least one feature derivation branch to determine a multidimensional state attribute vector set of the sensor measurement state data, wherein the feature granularity of the multidimensional state attribute vector set is smaller than that of the sensor measurement state data.

[0033] In one implementation, the multidimensional state attribute vector set, the target performance feature recognition result, and the target calibration feature recognition result are generated through at least one long short-term memory model, which is debugged through the following steps:

[0034] Determine a set of debugging samples, wherein the set of debugging samples includes at least one sample of sensor measurement state data, and the at least one sample of sensor measurement state data carries prior distribution information and prior bias information of the prior data to be adjusted;

[0035] For each sample of sensor measurement state data:

[0036] Determine the multidimensional state attribute vector set sample of the sensor measurement state data sample;

[0037] Based on the multidimensional state attribute vector set examples, determine the target performance feature identification result examples;

[0038] Based on the multidimensional state attribute vector set examples, determine the target proofreading feature recognition result examples;

[0039] Based on the target performance feature identification result examples and the target calibration feature identification result examples, determine the X-band performance feature examples and the horizontal correction expected feature examples;

[0040] Based on the X-band performance characteristic sample and the horizontal correction expected characteristic sample, the distribution information sample and deviation information sample of the data sample to be adjusted are determined in the sensing measurement state data sample;

[0041] Optimize the model weights of the at least one long short-term memory model so that the differences between the distribution information samples and deviation information samples of the data samples to be adjusted in the sensing measurement state data samples and the prior distribution information and prior deviation information of the prior data to be adjusted converge.

[0042] In one implementation, the difference includes at least one of the following comparison errors: distribution comparison error between the distribution information sample and the prior distribution information; and deviation comparison error between the deviation information sample and the prior deviation information.

[0043] This application provides an AI correction system, including at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to perform the above-described method.

[0044] This application provides a readable storage medium on which a program or instruction is stored, and when the program or instruction is executed by a processor, it implements the steps of the above method.

[0045] This application proposes a radar antenna array calibration scheme based on a multi-dimensional state attribute vector set and machine learning algorithm. By deeply analyzing sensor measurement state data, it can accurately identify the X-band performance characteristics of the antenna array and provide specific correction decision guidance for each performance characteristic. This approach enables intelligent and automated calibration of radar antenna arrays, significantly improving calibration accuracy and efficiency, and providing a more reliable guarantee for the widespread application of radar systems. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating an intelligent horizontal correction method for an X-band radar antenna provided in an embodiment of this application.

[0047] Figure 2 This is a schematic diagram of the structure of an AI correction system provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, in this application, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects have an "or" relationship.

[0050] Figure 1 A method for intelligent horizontal correction of an X-band radar antenna is shown, which is applied to an AI correction system. The method includes the following steps 110-150.

[0051] In modern radar systems, X-band radar is highly favored due to its powerful penetration and high resolution. However, to ensure the continuous and efficient operation of such radar systems, precise horizontal calibration of their antenna arrays is crucial. Traditional calibration methods often rely on manual intervention and regular maintenance, which is not only time-consuming and labor-intensive but may also introduce errors due to human factors. Therefore, the AI ​​calibration system in this application embodiment can automatically analyze electronic level data, identify and predict antenna horizontal deviations, thereby guiding the system to make precise adjustments. The implementation process of this application embodiment is described in detail below through steps 110-150 and related application scenario examples.

[0052] Step 110: Obtain the sensing measurement state data of the target radar antenna array and determine the multidimensional state attribute vector set of the sensing measurement state data.

[0053] For example, the AI ​​correction system first acquires real-time sensor measurement status data of the target radar antenna array using sensors and measurement devices built into the radar system. This sensor measurement status data covers multiple key parameters of the antenna array, including but not limited to pointing angle, signal strength, phase difference, ambient temperature, humidity, and possible mechanical vibrations. This sensor measurement status data is organized into a multi-dimensional state attribute vector set, providing a foundation for subsequent analysis and processing.

[0054] Step 120: Determine the target performance feature identification result based on the multidimensional state attribute vector set, wherein the target performance feature identification result includes several first antenna performance feature units, and the feature weight of each first antenna performance feature unit indicates the confidence level that the first antenna performance feature unit is an X-band performance feature.

[0055] Once sufficient sensor measurement state data has been collected, the AI ​​calibration system begins analyzing this data to determine the target performance characteristic identification results. This process involves in-depth mining of a multi-dimensional set of state attribute vectors to identify key features reflecting the antenna array's performance. These features are divided into several first antenna performance characteristic units, each representing a specific X-band performance characteristic, such as signal transmission efficiency, phase coherence, or pointing accuracy. Each first antenna performance characteristic unit is assigned a feature weight, which indicates the confidence level of that characteristic unit as an X-band performance characteristic, i.e., its importance in the overall performance evaluation.

[0056] Step 130: Determine the target calibration feature recognition result based on the multidimensional state attribute vector set, wherein the target calibration feature recognition result includes several second antenna performance feature units, each second antenna performance feature unit corresponds to a first antenna performance feature unit, and the feature weight of each second antenna performance feature unit indicates the calibration decision feature matched by the first antenna performance feature unit corresponding to that second antenna performance feature unit.

[0057] After identifying the antenna performance characteristics, the AI ​​calibration system further analyzes this data to determine the target calibration feature identification results. This process aims to find the calibration decision features corresponding to each first antenna performance feature element. Therefore, each first antenna performance feature element corresponds to a second antenna performance feature element, which contains the calibration strategy information required to achieve performance optimization. Similarly, each second antenna performance feature element is also assigned a feature weight that reflects the effectiveness and applicability of the corresponding calibration decision feature.

[0058] Step 140: Determine the X-band performance characteristics and the expected horizontal correction characteristics based on the target performance characteristic identification results and the target calibration characteristic identification results.

[0059] Based on the target performance feature identification results and the target calibration feature identification results, the AI ​​calibration system can comprehensively evaluate the current performance status of the antenna array and determine the key X-band performance characteristics and the required horizontal calibration expectation characteristics. This step is the bridge connecting performance analysis and actual calibration actions. It ensures that the calibration strategy can specifically address the specific problems existing in the antenna array, thereby improving the accuracy and efficiency of calibration.

[0060] Step 150: Determine the distribution information and deviation information of the data to be adjusted in the sensing measurement state data based on the X-band performance characteristics and the horizontal correction expectation characteristics, and calibrate the target radar antenna array based on the distribution information and the deviation information.

[0061] Finally, based on the determined X-band performance characteristics and desired horizontal correction characteristics, the AI ​​calibration system further analyzes the distribution and deviation information of the data to be adjusted in the sensor measurement state data. This information provides the system with clear guidance on how to adjust the antenna array to achieve optimal performance. Based on these analyses, the AI ​​calibration system can automatically execute calibration procedures, adjusting the antenna array's control parameters, such as phase control codes and phase shifter settings, to achieve precise horizontal correction. After calibration, the AI ​​calibration system also verifies the correction results to ensure that the antenna array's performance has been effectively improved.

[0062] Taking airport surface surveillance radar as an example, this type of radar typically needs to operate continuously in complex electromagnetic environments and variable weather conditions to ensure real-time monitoring of critical areas such as airport runways and aprons. However, due to environmental factors (such as temperature fluctuations and wind shear) and mechanical wear, the horizontal pointing accuracy of the radar antenna may gradually decrease, thus affecting the accuracy and reliability of the surveillance.

[0063] In this context, an AI correction system is integrated into the radar control software to achieve intelligent horizontal correction of the antenna array. The AI ​​correction system first collects real-time sensor measurement data of the antenna array using built-in sensors and measurement devices, including pointing angle, signal strength, phase difference, and ambient temperature and humidity. Then, the AI ​​correction system uses machine learning algorithms to analyze this data, identifying key characteristics reflecting antenna performance, such as signal transmission efficiency and phase consistency.

[0064] Next, the AI ​​calibration system determines the corresponding calibration decision features based on these performance characteristics, such as adjusting the phase control code to improve phase consistency or adjusting the antenna pointing angle to optimize signal transmission efficiency. After determining the key X-band performance characteristics and the desired horizontal calibration features, the AI ​​calibration system begins the calibration procedure. It automatically adjusts the antenna array's control parameters to achieve precise horizontal calibration and monitors performance changes in real time during the calibration process to ensure the accuracy and stability of the calibration results.

[0065] By introducing an AI-powered calibration system, the horizontal calibration process for airport surface surveillance radar antennas has achieved significant intelligence and automation. The AI ​​system can automatically identify and predict antenna horizontal deviations, guiding precise adjustments and significantly improving calibration efficiency and accuracy. Simultaneously, by reducing human intervention, the operational costs of the AI ​​system have been effectively reduced, providing a more reliable guarantee for the safe operation of the airport.

[0066] In conclusion, the application of the AI ​​correction system in intelligent horizontal correction of X-band radar antennas demonstrates its enormous potential and value. By integrating advanced AI algorithms and hardware acceleration technologies, this AI correction system can achieve efficient and accurate correction of antenna arrays, providing strong support for the widespread application of airport surface surveillance radar.

[0067] It is worth mentioning that steps 120 to 130 aim to accurately identify the performance characteristics of the target radar antenna array by deeply analyzing and processing the multi-dimensional state attribute vector set, and further determine the corresponding correction strategy.

[0068] First, in step 120, the AI ​​correction system uses advanced machine learning algorithms to perform in-depth mining and analysis based on the collected multi-dimensional state attribute vector set. During this process, the AI ​​correction system identifies several key first antenna performance characteristic units, which comprehensively reflect the antenna array's performance in the X-band. To quantify the importance of each characteristic unit in the overall performance evaluation, the AI ​​correction system assigns a feature weight to each first antenna performance characteristic unit. This weight essentially indicates the confidence level of the characteristic unit as an X-band performance feature, i.e., its accuracy and reliability in describing antenna performance.

[0069] Secondly, in step 130, the AI ​​correction system further utilizes a multi-dimensional state attribute vector set to determine the target correction feature recognition result. The core of this step is to find a corresponding second antenna performance feature element for each identified first antenna performance feature element. This corresponding second antenna performance feature element actually contains the correction strategy information required to achieve performance optimization, guiding the AI ​​correction system on how to adjust the antenna array to improve its performance. To evaluate the effectiveness of each correction strategy, the AI ​​correction system assigns a feature weight to each second antenna performance feature element. This weight reflects the effectiveness and applicability of the corresponding correction decision feature, i.e., the probability of successfully improving antenna performance in practical applications.

[0070] Through these two steps, the AI ​​correction system can not only accurately identify the key performance characteristics of the antenna array in the X-band, but also provide specific, quantified correction strategy guidance for each performance characteristic. This in-depth analysis and processing based on multi-dimensional state attribute vector sets ensures the pertinence and effectiveness of the correction strategy, thereby significantly improving the performance correction efficiency and accuracy of the antenna array.

[0071] Building on this, step 150 focuses on how to utilize the identified X-band performance characteristics and horizontal correction expectation characteristics to accurately calibrate the target radar antenna array.

[0072] First, the AI ​​calibration system delves into the sensor measurement state data based on the identified X-band performance characteristics and the desired horizontal calibration features. During this process, the AI ​​calibration system pays particular attention to data points closely related to the performance characteristics and calibration expectations to reveal the distribution and bias information of the data to be calibrated. The distribution information describes the distribution of the data in multidimensional space, while the bias information reflects the difference between the current antenna array performance and the desired performance.

[0073] Secondly, the AI ​​calibration system formulates targeted calibration strategies based on the mined distribution and deviation information. This strategy aims to gradually reduce deviations by adjusting antenna array control parameters, such as phase control codes and phase shifter settings, so that the antenna array's performance gradually approaches the desired level. When formulating the calibration strategy, the AI ​​calibration system comprehensively considers various factors, such as the feasibility of the adjustment, the impact on other parts of the AI ​​calibration system, and the timeliness of the calibration, to ensure the smooth progress of the calibration process.

[0074] Finally, the AI ​​calibration system executes the calibration strategy, performing actual calibration on the target radar antenna array. During the calibration process, the AI ​​calibration system monitors the performance changes of the antenna array in real time to ensure the accuracy and stability of the calibration results. Once calibration is complete, the system verifies the calibration results, evaluating the effectiveness of the calibration by comparing them with the expected performance. If the calibration results meet expectations, the AI ​​calibration system ends the calibration process; if they do not meet expectations, the AI ​​calibration system adjusts the calibration strategy according to the actual situation and continues to execute the calibration operation until a satisfactory calibration effect is achieved.

[0075] As can be seen, the technical solution described in step 150 achieves precise calibration of the target radar antenna array by deeply mining the sensor measurement status data and formulating and implementing a targeted calibration strategy. This not only improves the efficiency and accuracy of calibration but also provides a strong guarantee for the stable operation of the radar system.

[0076] This application embodiment achieves a comprehensive and in-depth analysis of the antenna array's performance by acquiring the sensor measurement state data of the target radar antenna array and constructing a multi-dimensional state attribute vector set. By further determining the target performance feature identification results and target calibration feature identification results, X-band performance characteristics can be accurately identified, and specific correction decision guidance can be provided for each performance feature. Based on these identification results, the distribution and deviation information of the data to be adjusted can be accurately determined, thereby achieving precise calibration of the target radar antenna array. In summary, this application embodiment significantly improves calibration efficiency and accuracy, reduces the need for manual intervention, and provides strong support for the stable operation of the radar system.

[0077] In some optional embodiments, determining the target performance feature identification result based on the multidimensional state attribute vector set includes: performing a feature embedding operation on the multidimensional state attribute vector set to determine a multidimensional state attribute vector set with completed feature embedding, wherein the multidimensional state attribute vector set with completed feature embedding includes a plurality of third antenna performance feature units; for each third antenna performance feature unit in the multidimensional state attribute vector set with completed feature embedding, determining the target pointing angle feature variable in the pointing dimension of the multidimensional state attribute vector set with completed feature embedding, determining the target phase angle feature variable in the phase dimension of the multidimensional state attribute vector set with completed feature embedding, and determining the feature weight of the first antenna performance feature unit in the target performance feature identification result based on the target pointing angle feature variable and the target phase angle feature variable.

[0078] Applying this embodiment, the process of determining the target performance feature recognition result based on the multidimensional state attribute vector set first involves performing a feature embedding operation on the multidimensional state attribute vector set. This aims to transform the original state attribute vectors, which may contain redundancy and noise, into more expressive and discriminative feature vectors. Through this feature embedding, the AI ​​correction system can extract key features closely related to antenna performance from high-dimensional data, laying a solid foundation for subsequent performance feature recognition.

[0079] For example, feature embedding is achieved by applying a series of mathematical transformations and mapping functions to a multidimensional set of state attribute vectors. These transformations and functions may be linear or nonlinear, depending on the characteristics of the data and the type of features required. During feature embedding, the AI ​​correction system attempts to retain the information most critical to antenna performance identification while removing or reducing interference from irrelevant or redundant information. After feature embedding, the multidimensional set of state attribute vectors is transformed into a new set of feature vectors containing several third antenna performance feature units.

[0080] Next, for each third antenna performance feature unit in the multidimensional state attribute vector set after feature embedding, the AI ​​correction system further determines its target pointing angle feature variable in the pointing dimension. The pointing dimension is a crucial aspect of radar antenna performance, reflecting the antenna beam's pointing capability in space. By determining the target pointing angle feature variable, the AI ​​correction system can assess the antenna's radiation intensity and pointing accuracy in a specific direction. This step may rely on advanced signal processing algorithms and pattern recognition techniques to ensure accurate estimation of the pointing angle.

[0081] Simultaneously, the AI ​​correction system also determines the target phase angle feature variable in the phase dimension of the multi-dimensional state attribute vector set after feature embedding. The phase dimension is another key performance indicator, reflecting the phase relationship between various elements in the antenna array. Correct phase relationships are fundamental to antenna beamforming and pointing control. Therefore, by determining the target phase angle feature variable, the AI ​​correction system can evaluate the phase consistency and phase control capability of the antenna array. This step also requires precise mathematical models and algorithms to ensure accurate measurement and calculation of the phase angle.

[0082] Finally, the AI ​​correction system determines the feature weights of the first antenna performance feature units in the target performance feature identification results based on the target pointing angle feature variables and the target phase angle feature variables. Feature weight is a quantitative indicator that reflects the confidence and importance of each first antenna performance feature unit in describing the antenna's X-band performance. By comprehensively considering feature variables in both the pointing and phase dimensions, the AI ​​correction system can more comprehensively evaluate the contribution of each performance feature unit to the overall performance and assign corresponding feature weights accordingly. This process may require the application of machine learning algorithms or statistical models to accurately calculate and allocate the feature weights.

[0083] As can be seen, the process of determining the target performance characteristics based on a multi-dimensional state attribute vector set is a comprehensive data analysis and processing workflow. Through feature embedding operations, determination of feature variables in the pointing and phase dimensions, and calculation and allocation of feature weights, the AI ​​correction system can achieve accurate identification and quantitative evaluation of the X-band performance of radar antennas. This technical solution provides important basis and support for subsequent radar antenna calibration and optimization. It not only improves the accuracy and efficiency of calibration but also lays a solid foundation for the stable operation and performance improvement of the radar system.

[0084] In the following steps, determining the target pointing angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the pointing dimension includes: traversing the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first pointing feature label of the multidimensional state attribute vector set based on the pointing dimension, and determining the maximum value among the traversed feature weights as the first target pointing angle feature variable; and / or traversing the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second pointing feature label of the multidimensional state attribute vector set based on the pointing dimension, and determining the maximum value among the traversed feature weights as the second target pointing angle feature variable. Determining the target phase angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the phase dimension includes: traversing the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first phase feature label of the multidimensional state attribute vector set based on the phase dimension. The feature weights of all third antenna performance feature units between the third antenna performance feature units and the second phase feature label of the multidimensional state attribute vector set are determined, and / or the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second phase feature label of the multidimensional state attribute vector set are determined based on the phase dimension, and the maximum value of the feature weights is determined as the second target phase angle feature variable. The feature weights of the first antenna performance feature unit in the target performance feature identification result are determined based on the target pointing angle feature variable and the target phase angle feature variable, including: weighting the first target pointing angle feature variable and / or the second target pointing angle feature variable with the first target phase angle feature variable and / or the second target phase angle feature variable to determine the feature weights of the first antenna performance feature unit corresponding to the third antenna performance feature unit in the target performance feature identification result.

[0085] Based on this embodiment, determining the target pointing angle feature variable in the pointing dimension of the multidimensional state attribute vector set after feature embedding is a crucial process. This involves a comprehensive examination of the feature weights between the third antenna performance feature unit and the pointing feature label in the multidimensional state attribute vector set. For example, the AI ​​correction system first iterates through the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first pointing feature label in the multidimensional state attribute vector set, based on the pointing dimension. This iteration process aims to find the third antenna performance feature unit that best matches the first pointing feature label, i.e., the unit whose feature weight reaches its maximum value. This maximum value is determined as the first target pointing angle feature variable, representing the strongest correlation between the third antenna performance feature unit and the first pointing feature label in that pointing dimension.

[0086] Simultaneously, and / or representing another possible scenario, the AI ​​correction system also iterates through the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second pointing feature label of the multi-dimensional state attribute vector set, based on the pointing dimension. Similarly, the purpose of this iteration is to find the third antenna performance feature unit that best matches the second pointing feature label, and the maximum value of its feature weight is determined as the second target pointing angle feature variable. This variable reflects the strongest correlation between the third antenna performance feature unit and the second pointing feature label in the pointing dimension.

[0087] After determining the target pointing angle feature variable, the AI ​​correction system then determines the target phase angle feature variable in the phase dimension of the multidimensional state attribute vector set with completed feature embedding. This process is similar to the processing in the pointing dimension, but focuses on the phase feature label. For example, based on the phase dimension, the AI ​​correction system iterates through the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first phase feature label of the multidimensional state attribute vector set, and determines the maximum value among the iterated feature weights as the first target phase angle feature variable. This variable represents the strongest correlation between the third antenna performance feature unit and the first phase feature label in that phase dimension.

[0088] Similarly, and / or indicates that the AI ​​correction system also considers another scenario: traversing the feature weights of all third antenna performance feature elements between the third antenna performance feature element and the second phase feature label of the multi-dimensional state attribute vector set based on the phase dimension. The purpose of this traversal is to find the third antenna performance feature element that best matches the second phase feature label, and the maximum value of its feature weight is determined as the second target phase angle feature variable. This variable reflects the strongest correlation between the third antenna performance feature element and the second phase feature label in the phase dimension.

[0089] Finally, based on the determined target pointing angle and phase angle feature variables, the AI ​​correction system further determines the feature weights of the first antenna performance feature unit in the target performance feature identification result. This step is achieved by weighting the first and / or second target pointing angle feature variables with the first and / or second target phase angle feature variables. The specific weighting method may depend on the actual application scenario and specific requirements, but the core purpose is to comprehensively consider the feature variables in the pointing and phase dimensions to determine the confidence and importance of the first antenna performance feature unit in the overall performance evaluation. Through this weighting process, the AI ​​correction system can derive the feature weights of the first antenna performance feature unit corresponding to the third antenna performance feature unit, thus providing important basis and support for subsequent radar antenna calibration and optimization.

[0090] This design, which identifies the target feature variables in the pointing and phase dimensions of the multidimensional state attribute vector set after feature embedding, and determines the feature weights in the target performance feature recognition result based on these variables, is a comprehensive data analysis and processing process. This process involves not only in-depth traversal and examination of the multidimensional state attribute vector set, but also comprehensive consideration of feature variables in different dimensions and their interrelationships. Through this technical solution, the AI ​​correction system can achieve accurate identification and quantitative evaluation of radar antenna performance, providing strong support for subsequent antenna calibration and optimization. This not only improves the accuracy and efficiency of calibration, but also lays a solid foundation for the stable operation and performance improvement of the radar system.

[0091] In some examples, the target performance feature identification result includes at least one performance attention index, where the feature weight of the first antenna performance feature element in each performance attention index indicates the confidence level that the first antenna performance feature element is an X-band performance feature of the target of the corresponding performance type.

[0092] Understandably, in practical applications, target performance feature identification results are a comprehensive and in-depth evaluation product. It not only encompasses multiple dimensions of radar antenna performance but also quantifies the characteristics and importance of these performance metrics through specific performance indicators. A key component of this is the performance attention index, which reveals different aspects of antenna performance characteristics in a unique way.

[0093] The performance attention metric in this embodiment is a quantitative measure used to represent the confidence level of the X-band performance characteristics exhibited by the first antenna performance feature element under a specific performance category. In this embodiment, "confidence level" is a commonly used term in statistics and machine learning, reflecting the degree of trust the AI ​​correction system has in a particular result or prediction. In radar antenna performance evaluation, a higher confidence level means that the AI ​​correction system is more confident in the antenna's performance in a specific performance context.

[0094] Based on this, the performance attention metric is constructed using feature embedding and feature weight calculation of the multidimensional state attribute vector set mentioned earlier. Specifically, when the AI ​​correction system performs feature embedding on the multidimensional state attribute vector set, it is actually transforming these high-dimensional raw data, which may contain redundancy and noise, into more expressive and discriminative feature vectors. Each element in these feature vectors can be considered a quantitative representation of a specific aspect of the antenna performance.

[0095] Next, after determining the multidimensional state attribute vector set for feature embedding, the AI ​​correction system further calculates the target feature variables for each third antenna performance feature element in the pointing and phase dimensions, such as the target pointing angle feature variable and the target phase angle feature variable. These feature variables actually describe the antenna's radiation intensity and pointing accuracy in a specific direction, as well as the phase relationship between the various elements in the antenna array. By calculating these feature variables, the AI ​​correction system can gain a more comprehensive understanding of the antenna's performance characteristics.

[0096] However, simply defining the antenna's performance characteristics is insufficient; the AI ​​calibration system also needs to determine the importance of these characteristics in the overall performance evaluation. This is where feature weight calculation comes in. By calculating the feature weights of each third antenna performance feature element, the AI ​​calibration system can quantify the confidence and importance of that element in describing the antenna's X-band performance. These feature weights essentially tell the AI ​​calibration system which performance characteristics are most critical for antenna performance evaluation and which can be ignored or are secondary considerations.

[0097] Finally, once the AI ​​correction system possesses these feature weights, it can construct a performance attention metric. Specifically, the AI ​​correction system maps the feature weights of each third antenna performance feature unit to the corresponding first antenna performance feature unit, thereby obtaining the feature weight of that first antenna performance feature unit. This feature weight is actually part of the performance attention metric; it represents the confidence level of that first antenna performance feature unit on the target X-band performance characteristics of the corresponding performance category.

[0098] Through this process, the AI ​​calibration system can obtain a target performance feature recognition result that includes multiple performance attention indicators. This result not only tells the AI ​​calibration system which performance aspects of the antenna are good, but also the importance of these performance aspects in the overall evaluation. This is invaluable for subsequent radar antenna calibration and optimization, as it helps the AI ​​calibration system to more accurately pinpoint problems and make more targeted optimizations and improvements.

[0099] As can be seen, the performance attention metric is a crucial concept, providing AI calibration systems with a novel method for quantifying and evaluating radar antenna performance. By calculating the feature weights of each first antenna performance feature element, the AI ​​calibration system can gain a more comprehensive understanding of the antenna's performance characteristics and more accurately assess the importance of these characteristics in overall performance. This provides strong support for subsequent radar antenna calibration and optimization.

[0100] In some alternative embodiments, determining the target calibration feature recognition result based on the multidimensional state attribute vector set includes: performing a feature embedding operation on the multidimensional state attribute vector set to obtain the target calibration feature recognition result, wherein for each performance category's X-band performance feature, the target calibration feature recognition result includes at least one performance attention index, and the feature weight of the second antenna performance feature unit of each performance attention index indicates that the first antenna performance feature unit corresponding to the second antenna performance feature unit serves as a performance correction reference under the set correction rule matched by the X-band performance feature of that performance category.

[0101] Based on this embodiment, an innovative technical solution based on a multi-dimensional state attribute vector set is proposed for radar antenna performance calibration and feature recognition. The core of this solution lies in the fact that by performing feature embedding operations on the multi-dimensional state attribute vector set, the AI ​​calibration system can obtain an information-rich and structurally compact target calibration feature recognition result. This result not only comprehensively reflects the performance characteristics of the radar antenna but also provides the AI ​​calibration system with a deeper understanding and guidance for antenna performance calibration through an innovative approach: the performance attention index.

[0102] First, the AI ​​calibration system requires feature embedding of the multidimensional state attribute vector set. This step is fundamental to the entire technical solution; its purpose is to transform the original high-dimensional state attribute vectors, which may contain redundancy and noise, into more expressive and discriminative feature vectors. Each element in these feature vectors can be considered a quantitative representation of a specific aspect of the antenna performance. Through feature embedding, the AI ​​calibration system can better capture and describe the antenna's performance characteristics, providing strong support for subsequent performance calibration and feature recognition.

[0103] After completing the feature embedding operation, the AI ​​calibration system obtains a feature vector set rich in antenna performance information. However, this feature vector set is still relatively abstract and difficult to interpret directly. To more intuitively understand and evaluate antenna performance, the AI ​​calibration system needs to further construct a performance attention metric.

[0104] The performance attention metric is a key innovation in the AI ​​correction system technical solution. It is essentially a quantitative measure used to represent the confidence level of the performance correction reference under a specific performance category, where the first antenna performance characteristic element corresponding to the second antenna performance characteristic element serves as the X-band performance characteristic matched with a set correction rule. In this application embodiment, the "confidence level" reflects the degree of trust the AI ​​correction system has in the performance correction reference, while the "correction rule" is a series of correction standards and strategies formulated by the AI ​​correction system based on the antenna's performance characteristics and actual needs.

[0105] Furthermore, the AI ​​correction system calculates the feature weight of each second antenna performance feature element for the X-band performance characteristics of each performance category. This feature weight essentially quantifies the importance of that second antenna performance feature element in describing antenna performance, as well as its confidence level as a performance correction reference.

[0106] When calculating feature weights, the AI ​​calibration system comprehensively considers multiple factors, including but not limited to the antenna's radiation intensity, pointing accuracy, and phase relationship. These factors are key to antenna performance and are the focus of the AI ​​calibration system during performance correction. By calculating the feature weights of each second antenna performance feature element, the AI ​​calibration system obtains a comprehensive and quantitative performance evaluation result. This result not only tells the AI ​​calibration system which performance aspects of the antenna are good or bad, but also the importance of these performance aspects in the overall evaluation and their confidence level as a performance correction reference.

[0107] Finally, once the AI ​​correction system possesses these feature weights, it can construct a performance attention metric. Specifically, the AI ​​correction system maps the feature weights of each second antenna performance feature unit to the corresponding first antenna performance feature unit, thereby obtaining the feature weight of that first antenna performance feature unit. This feature weight is actually part of the performance attention metric; it represents the confidence level of the performance correction reference of that first antenna performance feature unit under the set correction rule matched to the X-band performance characteristics of the corresponding performance type.

[0108] Through this process, the AI ​​calibration system obtains a target calibration feature recognition result that includes multiple performance attention metrics. This result not only comprehensively reflects the antenna's performance characteristics but also provides the AI ​​calibration system with a deep understanding and guidance for antenna performance calibration through performance attention metrics. The AI ​​calibration system can then formulate targeted calibration strategies and optimization schemes based on these metrics, thereby more effectively improving the antenna's performance and stability.

[0109] Thus, by constructing feature embedding operations and performance attention metrics, a novel method for radar antenna performance calibration and feature recognition is provided for AI correction systems. This not only improves the accuracy and comprehensiveness of performance evaluation but also offers new ideas and directions for better understanding and improving radar antenna performance.

[0110] In another alternative embodiment, determining the X-band performance characteristics and the desired horizontal correction characteristics based on the target performance characteristic identification results and the target calibration characteristic identification results includes: determining a discrimination score for each first antenna performance characteristic unit belonging to the X-band performance characteristics based on the feature weight of each first antenna performance characteristic unit in the target performance characteristic identification results; identifying first antenna performance characteristic units with discrimination scores greater than a set discrimination score as the X-band performance characteristics; and determining the desired horizontal correction characteristics corresponding to the X-band performance characteristics in the target calibration characteristic identification results based on the distribution characteristics of the X-band performance characteristics in the target performance characteristic identification results.

[0111] It is understood that the task of this embodiment is to determine the X-band performance characteristics and corresponding horizontal correction expectation characteristics based on the target performance characteristic identification results and the target calibration characteristic identification results. The core of this task is to accurately identify which performance characteristics belong to the X-band from a large number of performance characteristics, and further find the horizontal correction expectation characteristics corresponding to these X-band performance characteristics, so as to provide guidance for subsequent radar antenna performance optimization and calibration.

[0112] First, the AI ​​correction system starts with the target performance feature identification results. These results constitute a complex dataset containing multiple first antenna performance feature elements and their feature weights. Each first antenna performance feature element represents a specific performance characteristic of the antenna, while the feature weight quantifies the importance of that characteristic in the overall performance evaluation. To determine which first antenna performance feature elements belong to the X-band performance characteristics, the AI ​​correction system needs to calculate a discrimination score for each element.

[0113] The calculation of the discrimination score is a process that comprehensively considers feature weights and performance category information. Specifically, the AI ​​correction system calculates a score that reflects the likelihood of the unit belonging to the X-band performance characteristics based on the feature weights of each first antenna performance feature unit and the distribution of these weights in the overall dataset. The higher the score, the more likely the unit is to be part of the X-band performance characteristics.

[0114] After calculating the discrimination score, the AI ​​correction system can determine which first antenna performance characteristic elements truly belong to the X-band performance characteristics based on the set discrimination score threshold. Specifically, the AI ​​correction system selects the first antenna performance characteristic elements with a discrimination score greater than the set threshold as representatives of the X-band performance characteristics. These selected elements are the performance characteristics that the AI ​​correction system needs to focus on in subsequent analysis.

[0115] However, simply identifying the X-band performance characteristics is insufficient; the AI ​​correction system also needs to find the corresponding horizontal correction expectation characteristics. This requires the AI ​​correction system to turn to the target correction feature recognition results. These target correction feature recognition results are also a dataset containing multiple performance feature units and their feature weights, but they focus more on information related to antenna performance correction.

[0116] In the target calibration feature recognition results, each performance feature unit has one or more corresponding calibration rules. These calibration rules are formulated based on the actual performance characteristics and calibration requirements of the antenna, and they describe how to calibrate the antenna performance to achieve the desired level. The task of the AI ​​calibration system is to find the desired level calibration feature corresponding to the X-band performance characteristics among these calibration rules.

[0117] To accomplish this task, the AI ​​correction system first needs to analyze the distribution characteristics of X-band performance features within the target performance feature identification results. This includes understanding the proportion of these features in the overall dataset, the correlations between them, and their importance in antenna performance evaluation. Through these analyses, the AI ​​correction system can gain a more comprehensive and in-depth understanding of the X-band performance features.

[0118] The AI ​​correction system then matches this information with the correction rules in the target correction feature recognition results. Specifically, the AI ​​correction system searches for correction rules that are closely related to the X-band performance characteristics and can effectively correct them. The performance characteristics described by these correction rules are the expected horizontal correction characteristics that the AI ​​correction system is looking for. They represent the ideal state or expected level that the AI ​​correction system hopes to achieve when correcting the antenna performance.

[0119] Through this process, the AI ​​correction system can accurately determine the X-band performance characteristics and corresponding expected horizontal correction characteristics based on the target performance feature recognition results and target calibration feature recognition results. This result not only provides the AI ​​correction system with crucial information about antenna performance but also offers strong guidance for subsequent radar antenna performance optimization and correction work. The AI ​​correction system can use this information to formulate targeted correction strategies and optimization schemes, thereby more effectively improving antenna performance and stability. Simultaneously, this result can also serve as an important reference in the antenna design and manufacturing process, helping to better understand and improve the performance characteristics of radar antennas.

[0120] Under some preferred technical approaches, determining the distribution information and deviation information of the data to be adjusted in the sensing measurement state data based on the X-band performance characteristics and the desired horizontal correction characteristics includes: determining the feature association description between the multidimensional state attribute vector set and the sensing measurement state data; determining the distribution information and initial noise information of the data to be adjusted in the sensing measurement state data based on the X-band performance characteristics, the desired horizontal correction characteristics, and the feature association description; determining the noise disturbance used for the initial noise information based on the multidimensional state attribute vector set; and determining the deviation information of the data to be adjusted in the sensing measurement state data based on the noise disturbance and the initial noise information.

[0121] Understandably, the challenge of this technical approach is to accurately determine the distribution and bias information of the data to be adjusted from the sensor measurement state data. Accurately determining the distribution and bias information of the data to be adjusted from the sensor measurement state data requires guidance from X-band performance characteristics and expected horizontal correction characteristics, while also fully utilizing the inherent relationship between the multi-dimensional state attribute vector set and the sensor measurement state data.

[0122] First, the AI ​​calibration system needs to determine the feature correlation description between the multidimensional state attribute vector set and the sensor measurement state data. This is a crucial preliminary step because it provides the basic framework for subsequent analysis by the AI ​​calibration system. The multidimensional state attribute vector set is a dataset containing multiple state attributes, each describing the state or characteristic of the radar antenna or related system in a certain aspect. The sensor measurement state data, on the other hand, is the actual measured data, reflecting the actual performance of the antenna or system under specific conditions. By deeply analyzing these two sets of data, the AI ​​calibration system can discover the inherent connections and patterns between them; these connections and patterns constitute the feature correlation description.

[0123] With the feature association description, the AI ​​correction system can further determine the distribution information and initial noise information of the data to be adjusted in the sensor measurement state data based on the X-band performance characteristics, the desired horizontal correction characteristics, and the feature association description. The X-band performance characteristics are the key performance characteristics that the AI ​​correction system focuses on, representing the antenna's critical performance in the X-band. The desired horizontal correction characteristics are the target state that the AI ​​correction system hopes to achieve when correcting the antenna performance. By combining these two characteristics and the feature association description, the AI ​​correction system can locate the data to be adjusted in the sensor measurement state data that are related to the X-band performance characteristics and the desired horizontal correction characteristics, and understand their distribution and initial noise level.

[0124] Next, the AI ​​calibration system needs to determine the noise perturbation used for initial noise information based on a multi-dimensional state attribute vector set. Noise perturbation is an important concept, representing random fluctuations in sensor measurement state data caused by various factors (such as measurement errors, environmental interference, etc.). By utilizing information from the multi-dimensional state attribute vector set, the AI ​​calibration system can more accurately estimate and model noise perturbation. Specifically, the AI ​​calibration system can analyze the relationship between each attribute in the multi-dimensional state attribute vector set and the noise in the sensor measurement state data, thereby identifying those attributes that have a significant impact on noise, and using this information to construct a model of the noise perturbation.

[0125] With a model of the noise disturbance, the AI ​​correction system can determine the deviation information of the data to be adjusted from the sensor measurement state data based on the noise disturbance and initial noise information. The deviation information reflects the degree of difference between the data to be adjusted and its true or expected value. By combining the noise disturbance and initial noise information, the AI ​​correction system can more accurately assess the deviation of the data to be adjusted. Specifically, the AI ​​correction system can use the noise disturbance model to simulate random fluctuations in the sensor measurement state data and compare these fluctuations with the initial noise information to derive the deviation information of the data to be adjusted.

[0126] Thus, by deeply analyzing the intrinsic relationship between the multidimensional state attribute vector set and the sensor measurement state data, and by fully utilizing the performance characteristics of the X-band and the expected characteristics of horizontal correction, the distribution and deviation information of the data to be adjusted were accurately determined from the sensor measurement state data. This result not only provides the AI ​​correction system with crucial information about antenna performance but also offers strong guidance for subsequent radar antenna performance optimization and correction. The AI ​​correction system can use this information to formulate targeted correction strategies and optimization schemes, thereby more effectively improving antenna performance and stability.

[0127] In other examples, the X-band performance characteristics and the horizontal correction expectation characteristics include X-band performance characteristics for the data to be adjusted for at least one time-series label and horizontal correction expectation characteristics for the data to be adjusted for that performance class.

[0128] In practical applications, it is crucial to process and analyze the X-band performance characteristics and horizontal correction expectations in a more specific and in-depth manner. These characteristics are not merely abstract descriptions or generalizations, but are specific to the data to be adjusted for each time series label, providing more refined guidance for the performance optimization and correction of radar antennas.

[0129] First, X-band performance characteristics refer to the performance of radar antennas in the X-band. Radar antennas typically operate in multiple frequency bands, with the X-band being a crucial component. X-band performance characteristics describe and quantify various performance indicators of the antenna in the X-band, such as gain, directivity, and beamwidth. These performance indicators directly affect the radar system's detection capability, positioning accuracy, and anti-jamming capability, making them key considerations in antenna design and optimization.

[0130] The desired horizontal correction characteristic is the target state that the AI ​​correction system expects to achieve when correcting the performance of a radar antenna. In practical applications, due to various factors, the performance of a radar antenna often deviates from its ideal state. To improve antenna performance, the AI ​​correction system needs to correct it to bring it as close as possible to or reach the performance level expected by the AI ​​correction system. The desired horizontal correction characteristic describes this expected state, providing the AI ​​correction system with the direction and target of correction.

[0131] Accordingly, the X-band performance characteristics and expected horizontal correction characteristics are further refined, down to the data to be adjusted for each time tag. A time tag is an important concept, representing the temporal order or time series of data. During the operation of the radar system, antenna performance data is continuously generated, and this data is recorded in chronological order to form time-series data. Each time tag corresponds to a specific set of performance data, reflecting the antenna's performance state at a certain moment or over a certain period.

[0132] Therefore, when it comes to X-band performance characteristics and horizontal correction expectation characteristics, which include the data to be adjusted for at least one time tag, the AI ​​correction system is actually saying that these characteristics are not just a general description of antenna performance, but specific to each time tag, providing corresponding characteristics and expectations for the performance data at each moment or time period.

[0133] This design firstly enhances the AI ​​calibration system's ability to understand and analyze antenna performance. By examining the performance and expected characteristics of each time stamp, the AI ​​calibration system can more clearly understand the antenna's performance at different times or under different conditions, and what kind of calibration and optimization is needed. Secondly, it provides the AI ​​calibration system with more refined calibration and optimization strategies. Since the data on each time stamp has its specific characteristics and expectations, the AI ​​calibration system can formulate more targeted calibration strategies and optimization schemes based on this data, thereby more effectively improving antenna performance.

[0134] Specifically, once the AI ​​calibration system possesses the X-band performance characteristics and desired horizontal calibration characteristics of the data to be calibrated for at least one time-series label, it can perform the following steps: First, the AI ​​calibration system can extract and analyze features from the performance data for each time-series label. This includes calculating various performance indicators, identifying performance patterns, and detecting performance anomalies. Through these operations, the AI ​​calibration system can gain a deeper understanding of the antenna's performance at different times or under different conditions, providing fundamental data for subsequent calibration and optimization. Second, the AI ​​calibration system can formulate calibration strategies based on the desired horizontal calibration characteristics. For the data at each time-series label, the AI ​​calibration system can determine the calibration target and direction based on its desired characteristics. For example, if the desired characteristic is to increase gain, the AI ​​calibration system can formulate corresponding strategies to adjust the antenna's parameters or structure to improve its gain performance in the X-band. Then, the AI ​​calibration system can implement the calibration strategy and verify the results. After implementing the calibration strategy, the AI ​​calibration system needs to remeasure and evaluate the antenna's performance to verify the effectiveness of the calibration strategy. If the calibration results meet expectations, the AI ​​calibration system can apply the calibration strategy to data on other time-series labels. If the results do not meet expectations, the AI ​​calibration system needs to adjust or reformulate the strategy. Finally, the AI ​​calibration system can integrate and analyze the calibrated data to evaluate the effectiveness of the entire calibration process. By comparing the performance data before and after calibration, the AI ​​calibration system can understand the extent to which the calibration strategy improves antenna performance and provide guidance for subsequent optimization efforts.

[0135] As can be seen, the X-band performance characteristics and expected horizontal correction characteristics are further refined to the data to be adjusted for each time-series label. This processing method improves the AI ​​correction system's understanding and analysis capabilities of antenna performance, providing it with more refined correction and optimization strategies. Through a series of operations including feature extraction, strategy formulation, implementation verification, and effect evaluation, the AI ​​correction system can more effectively improve the performance of radar antennas and provide strong support for their performance in practical applications.

[0136] In some alternative examples, the sensing measurement state data is radar antenna array state data collected by a built-in measurement device. Before determining the multidimensional state attribute vector set of the sensing measurement state data, the method further includes: adjusting according to the original feature values ​​of each antenna performance feature element in the sensing measurement state data.

[0137] Based on this embodiment, especially when the data comes from radar antenna array status data collected by the built-in measurement device, the data processing is not limited to simple collection and organization, but involves more complex and refined feature adjustment and construction of multi-dimensional state attribute vector sets.

[0138] First, the sensor measurement status data is directly acquired by the built-in measurement equipment, reflecting various states of the radar antenna array during actual operation. This data is raw and unprocessed, and therefore may contain various noises, errors, and redundant information. In order to extract useful information from this raw data, the AI ​​correction system needs to perform a series of processing and analysis tasks.

[0139] Before determining the multidimensional state attribute vector set of the sensor measurement state data, an important step is to adjust the original eigenvalues ​​of each antenna performance characteristic element in the sensor measurement state data. This step is crucial because it directly affects the accuracy and effectiveness of the subsequent multidimensional state attribute vector set.

[0140] Raw eigenvalues ​​are the most fundamental elements in sensor measurement state data; they represent the original measured or calculated values ​​of antenna performance characteristic elements in a certain aspect. However, these raw eigenvalues ​​are often not the final state attributes required by the AI ​​correction system, as they may deviate from the true values ​​due to various factors. Therefore, the AI ​​correction system needs to adjust these raw eigenvalues ​​to eliminate or reduce the influence of these factors.

[0141] Adjusting the original feature values ​​can be a complex and delicate operation. It may involve multiple aspects such as data filtering, denoising, correction, and normalization. Filtering aims to eliminate random noise and interference in the data, making it smoother and more stable; denoising removes outliers and erroneous values ​​to avoid their impact on subsequent analysis; correction adjusts biases and distortions in the data to make it closer to the true values; and normalization transforms the data to a uniform scale for easier subsequent processing and analysis.

[0142] In adjusting the original feature values, the AI ​​correction system also needs to fully utilize prior knowledge and empirical data. Prior knowledge refers to the AI ​​correction system's understanding and knowledge of the radar antenna performance characteristics, including the changing patterns, influencing factors, and correlations of performance characteristics. Empirical data refers to the data and experience accumulated by the AI ​​correction system in processing and analyzing similar antenna performance characteristics in the past. By combining prior knowledge and empirical data, the AI ​​correction system can more accurately adjust the original feature values ​​to better reflect the actual antenna performance characteristics.

[0143] After adjusting the original feature values, the AI ​​calibration system can further construct a multidimensional state attribute vector set of the sensor measurement state data. A multidimensional state attribute vector set is a dataset containing multiple state attributes, each describing the state or characteristic of the radar antenna or related system in a certain aspect. By constructing this multidimensional state attribute vector set, the AI ​​calibration system can gain a more comprehensive and in-depth understanding of the state and performance characteristics of the radar antenna array.

[0144] When constructing a multidimensional state attribute vector set, the AI ​​correction system needs to select appropriate state attributes based on actual requirements and application scenarios. These state attributes should comprehensively reflect the state and performance characteristics of the radar antenna array, while also being measurable and analyzable. For example, the AI ​​correction system can choose antenna gain, directivity, beamwidth, VSWR, etc., as state attributes, because they are all important indicators describing antenna performance.

[0145] Simultaneously, the AI ​​correction system also needs to quantify and standardize the selected state attributes. Quantization refers to converting the raw values ​​of the state attributes into numerical forms for subsequent calculations and analysis; standardization aims to eliminate differences in dimensions and scales between different state attributes, making them comparable and additive. Through quantization and standardization, the AI ​​correction system can transform the selected state attributes into elements of a multi-dimensional state attribute vector set, providing a foundation for its subsequent analysis and applications.

[0146] It is evident that when the sensor measurement state data is radar antenna array state data collected by built-in measurement equipment, the AI ​​correction system needs to adjust the original feature values ​​of each antenna performance characteristic unit in the sensor measurement state data before determining its multidimensional state attribute vector set. This step is crucial, as it directly affects the accuracy and effectiveness of the subsequent multidimensional state attribute vector set. By adjusting the original feature values, constructing the multidimensional state attribute vector set, and selecting appropriate state attributes for quantification and standardization, the AI ​​correction system can gain a more comprehensive and in-depth understanding of the state and performance characteristics of the radar antenna array, providing a solid foundation for subsequent analysis and applications.

[0147] In some preferred embodiments, determining the multidimensional state attribute vector set of the sensing measurement state data includes: processing the sensing measurement state data through at least one feature embedding branch to obtain sensing measurement state data with completed feature embedding; performing at least one feature pooling and at least one feature derivation on the sensing measurement state data with completed feature embedding through at least one feature pooling branch and at least one feature derivation branch to determine the multidimensional state attribute vector set of the sensing measurement state data, wherein the feature granularity of the multidimensional state attribute vector set is smaller than that of the sensing measurement state data.

[0148] This embodiment focuses on how to determine the multidimensional state attribute vector set of sensor measurement state data. This process involves not only data processing and analysis, but also several key steps such as feature embedding, feature pooling, and feature derivation.

[0149] First, to determine the multidimensional state attribute vector set of the sensor measurement state data, the AI ​​calibration system needs to effectively process and analyze this raw data. Feature embedding is the first step in this process. The purpose of feature embedding is to transform the high-dimensional and complex information in the sensor measurement state data into a more compact, low-dimensional, and easily processed feature representation. This step is usually achieved through a feature embedding branch, which can perform deep feature extraction and transformation on the raw data, thereby obtaining the sensor measurement state data with complete feature embedding. This data processed by feature embedding not only retains the key information in the original data but also facilitates subsequent feature pooling and feature derivation operations.

[0150] Secondly, after feature embedding, the AI ​​correction system needs to perform further feature pooling and feature derivation processing on the obtained sensor measurement state data. Feature pooling is an effective dimensionality reduction technique that reduces data dimensionality by merging or aggregating similar features while retaining key information. This step is usually performed by the feature pooling branch, which further compresses and refines the feature-embedded data to obtain a more compact and informative feature representation. Feature derivation, on the other hand, generates new features based on existing features through certain mathematical transformations or combinations. These new features often reveal hidden information or patterns in the data. Feature derivation is usually implemented by the feature derivation branch, which further mines and explores the pooled data to obtain richer and more diverse feature representations.

[0151] Through feature pooling and feature derivation, the AI ​​correction system can obtain a series of feature representations with different granularities and levels. These feature representations together constitute a multidimensional state attribute vector set of the sensor measurement state data. It is worth noting that the feature granularity of this multidimensional state attribute vector set is smaller than that of the original sensor measurement state data. This means that the AI ​​correction system, through a series of operations such as feature embedding, feature pooling, and feature derivation, successfully transforms high-dimensional, complex sensor measurement state data into a low-dimensional, compact, and information-rich multidimensional state attribute vector set.

[0152] The advantage of this technical solution lies in its ability to effectively process and analyze sensor measurement state data, extract key feature information, and construct a multi-dimensional state attribute vector set with rich representational capabilities. This not only improves data processing efficiency and analytical accuracy but also provides strong data support for subsequent tasks such as machine learning and pattern recognition. Furthermore, through the organic combination of steps such as feature embedding, feature pooling, and feature derivation, the AI ​​correction system can more flexibly adjust and optimize the construction process of the multi-dimensional state attribute vector set to adapt to different application scenarios and practical needs.

[0153] In practical implementation, the AI ​​correction system can select appropriate feature embedding, feature pooling, and feature derivation branches based on the characteristics and analytical needs of the sensor measurement state data. For example, for sensor measurement state data containing time-series data, the AI ​​correction system can choose a feature embedding branch based on recurrent neural networks to extract dynamic features from the time series; for sensor measurement state data containing image data, the AI ​​correction system can choose a feature embedding branch based on convolutional neural networks to extract spatial features from the image. Similarly, the AI ​​correction system can select appropriate feature pooling methods and feature derivation strategies according to actual needs to construct a multi-dimensional state attribute vector set best suited to the current application scenario.

[0154] This design allows the AI ​​correction system to determine a multidimensional set of state attribute vectors from sensor measurement data through a series of operations, including feature embedding, feature pooling, and feature derivation. This technical solution not only improves the efficiency and accuracy of data processing and analysis but also provides strong data support for subsequent machine learning, pattern recognition, and other tasks. By flexibly selecting and combining different feature embedding, feature pooling, and feature derivation branches, the AI ​​correction system can adapt to different application scenarios and practical needs, constructing a multidimensional set of state attribute vectors with rich representational capabilities.

[0155] In some preferred embodiments, the multidimensional state attribute vector set, the target performance feature identification result, and the target calibration feature identification result are generated by at least one long short-term memory model. The at least one long short-term memory model is debugged through the following steps: determining a debugging sample set, wherein the debugging sample set includes at least one sensor measurement state data sample, the at least one sensor measurement state data sample carrying prior distribution information and prior bias information of the prior data to be adjusted; for each sensor measurement state data sample: determining a multidimensional state attribute vector set sample for that sensor measurement state data sample; determining a target performance feature identification result sample based on the multidimensional state attribute vector set sample; determining a target calibration feature identification result sample based on the multidimensional state attribute vector set sample; and determining a target calibration feature identification result sample based on the target performance feature identification result. The X-band performance feature sample and the horizontal correction expected feature sample are determined based on the sample results and the target calibration feature recognition result. Distribution information samples and deviation information samples of the data sample to be adjusted are determined in the sensor measurement state data sample based on the X-band performance feature sample and the horizontal correction expected feature sample. The model weights of the at least one long short-term memory model are optimized so that the differences between the distribution information samples and deviation information samples of the data sample to be adjusted in the sensor measurement state data sample and the prior distribution information and prior deviation information of the prior data to be adjusted converge. For example, the differences include at least one of the following comparison errors: distribution comparison error between the distribution information sample and the prior distribution information; and deviation comparison error between the deviation information sample and the prior deviation information.

[0156] In practical applications, the generation of multidimensional state attribute vector sets, target performance feature recognition results, and target calibration feature recognition results are all implemented using at least one long short-term memory model. To ensure the effectiveness and accuracy of this model, a series of adjustments were made.

[0157] First, a debug sample set needs to be constructed, which is the cornerstone of model debugging. This debug sample set is not randomly combined, but carefully selected, and contains at least one sensor measurement state data sample. Each sensor measurement state data sample is like a treasure trove, containing rich prior distribution information and prior bias information of the prior data to be adjusted. This information is crucial for subsequent model debugging and optimization; it can guide the AI ​​calibration system to find the correct direction for model debugging.

[0158] Next, the AI ​​calibration system performs in-depth analysis and processing on each sensor measurement state data sample. The AI ​​calibration system first determines a multi-dimensional state attribute vector set sample for each sensor measurement state data sample. This step extracts key features from the data, laying the foundation for subsequent performance feature identification and calibration feature identification. With the multi-dimensional state attribute vector set sample, the AI ​​calibration system can further determine the target performance feature identification result sample and the target calibration feature identification result sample. These two result samples are important references for the AI ​​calibration system to debug its model, helping it evaluate the model's recognition and calibration capabilities.

[0159] However, the debugging process of the AI ​​calibration system does not stop there. Based on the target performance feature identification results and target calibration feature identification results, the AI ​​calibration system also needs to further determine X-band performance feature samples and horizontal calibration expectation feature samples. These two samples are the ultimate goals of the AI ​​calibration system's debugging model; they will guide the AI ​​calibration system to optimize the model's weights, enabling the model to more accurately identify and calibrate key features in the sensor measurement state data.

[0160] Next, optimizing the model's weights requires the AI ​​calibration system to determine the distribution information samples and bias information samples of the data samples to be adjusted based on the performance of X-band performance characteristic samples and horizontal correction expectation characteristic samples in the sensor measurement state data samples. These two information samples are important indicators for the AI ​​calibration system to evaluate model performance, and they will help the AI ​​calibration system discover deficiencies in the model during the identification and calibration process.

[0161] Finally, the AI ​​calibration system enters the model weight optimization stage. The goal of the AI ​​calibration system is to converge the differences between the distribution and bias information of the data samples to be adjusted in the sensor measurement state data samples and the prior distribution and bias information of the prior data to be adjusted. In other words, the AI ​​calibration system aims to make the model's performance during the recognition and calibration process closer to the AI ​​calibration system's expectations, making the model's output more accurate and reliable. To achieve this goal, the AI ​​calibration system continuously optimizes the model's weights until they meet the AI ​​calibration system's requirements.

[0162] During the optimization process, the AI ​​correction system pays special attention to the distribution comparison error between the distribution information samples and the prior distribution information, as well as the deviation comparison error between the deviation information samples and the prior deviation information. These two comparison errors are key indicators for the AI ​​correction system to evaluate model performance. They help the AI ​​correction system identify problems in the model's recognition and correction process and guide the AI ​​correction system to make targeted optimizations and improvements.

[0163] As can be seen, the AI ​​correction system generates a multi-dimensional state attribute vector set, target performance feature recognition results, and target calibration feature recognition results through at least one long short-term memory model. To ensure the effectiveness and accuracy of the model, the AI ​​correction system underwent a series of debugging steps. The system constructed a debugging sample set, extracted multi-dimensional state attribute vector set samples, determined target performance feature recognition result samples and target calibration feature recognition result samples, and further determined X-band performance feature samples and horizontal correction expected feature samples. Then, the AI ​​correction system optimized the model's weights based on these samples, making the model's performance in the recognition and calibration process closer to the AI ​​correction system's expectations. This series of debugging and optimization steps ensures that the AI ​​correction system's model can achieve optimal performance in practical applications.

[0164] In addition, in conjunction with the foregoing, for some independently implementable technical solutions, the calibration of the target radar antenna array based on the distribution information and the deviation information includes: obtaining a global influence information set of the target radar antenna array based on the distribution information and the deviation information, wherein the global influence information set includes Q uninterrupted global influence information, where Q is an integer greater than or equal to 1; obtaining a mechanical and electronic influence information set based on the global influence information set, wherein the mechanical and electronic influence information set includes Q uninterrupted mechanical and electronic influence information; and obtaining a global influence factor relationship vector network based on the global influence information set through a first relationship vector identification sub-algorithm included in the calibration control algorithm. The system comprises a set of global influencing factor relationship vector networks, wherein the set includes Q global influencing factor relationship vector networks; based on the set of mechanical and electronic influence information, the system obtains the set of mechanical and electronic influence factor relationship vector networks through a second relationship vector identification sub-algorithm included in the calibration control algorithm, wherein the set of mechanical and electronic influence factor relationship vector networks includes Q mechanical and electronic influence factor relationship vector networks; based on the set of global influencing factor relationship vector networks and the set of mechanical and electronic influence factor relationship vector networks, the system obtains the radar antenna array calibration label corresponding to the global influence information set through a calibration execution matching sub-algorithm included in the calibration control algorithm; and performs dynamic calibration on the target radar antenna array according to the radar antenna array calibration label.

[0165] Understandably, the process of calibrating a target radar antenna array must consider not only the physical characteristics of the radar antenna array itself, but also various influencing factors in its actual operating environment.

[0166] First, a global impact information set for the target radar antenna array needs to be obtained based on the distribution information and the deviation information. This step is fundamental to the calibration process because the global impact information set contains the influence of various external and internal factors that the radar antenna array may be affected by during operation. These factors collectively constitute the global operating environment of the radar antenna array. The global impact information set includes Q uninterrupted global impact information, where Q is an integer greater than or equal to 1. This means that the AI ​​correction system considers the influencing factors continuously and comprehensively, without omitting any factor that may affect the performance of the radar antenna array.

[0167] Next, the AI ​​correction system needs to acquire the electromechanical influence information set based on the global influence information set. The electromechanical influence information set is an important subset of the global influence information set, focusing on influencing factors directly related to the mechanical and electronic components of the radar antenna array. These influencing factors include, but are not limited to, component wear, aging, and electromagnetic interference, which have a direct and significant impact on the performance of the radar antenna array. The electromechanical influence information set also includes Q continuous electromechanical influence information sets to ensure that the AI ​​correction system has a comprehensive understanding and grasp of every electromechanical factor that may affect the performance of the radar antenna array.

[0168] With the global impact information set and the mechatronics impact information set, the AI ​​calibration system can further obtain the global impact factor relationship vector network set and the mechatronics impact factor relationship vector network set through calibration control algorithms. These two sets are crucial in the calibration process; they describe and represent the interactions and relationships between various impact factors through complex relationship vector networks. The global impact factor relationship vector network set includes Q global impact factor relationship vector networks, each corresponding to a global impact information, used to reveal the intrinsic connections between this impact information and other impact information. Similarly, the mechatronics impact factor relationship vector network set also includes Q mechatronics impact factor relationship vector networks, used to reveal the interactions and relationships between mechatronics impact information.

[0169] After acquiring the global influencing factor relationship vector set and the mechanical and electronic influencing factor relationship vector set, the AI ​​calibration system can obtain the radar antenna array calibration tag corresponding to the global influencing information set through the calibration execution matching sub-algorithm included in the calibration control algorithm. This calibration tag is the direct output of the calibration process; it contains all the necessary calibration information and instructions to guide subsequent calibration operations. The acquisition of the calibration tag is based on in-depth analysis and comprehensive judgment of the global influencing factor relationship vector set and the mechanical and electronic influencing factor relationship vector set, ensuring the accuracy and effectiveness of the calibration tag.

[0170] Finally, the AI ​​calibration system can perform dynamic calibration of the target radar antenna array based on the radar antenna array calibration label. Dynamic calibration is the final and one of the most important steps in the calibration process. It requires the AI ​​calibration system to perform real-time, dynamic calibration operations on the radar antenna array in the actual operating environment to ensure that the radar antenna array is always in optimal working condition. The dynamic calibration process is complex and precise. It requires the AI ​​calibration system to make full use of the rich information and data acquired earlier, such as the global influence information set, the mechanical and electronic influence information set, the global influence factor relationship vector network set, and the mechanical and electronic influence factor relationship vector network set, to eliminate the adverse effects of various influencing factors on the performance of the radar antenna array through precise calibration operations and adjustments, thereby ensuring the accuracy and stability of the radar antenna array.

[0171] As can be seen, calibrating a target radar antenna array is a comprehensive and complex process. It requires the AI ​​calibration system to fully consider various influencing factors in the actual operating environment of the radar antenna array. This involves comprehensively understanding and mastering these influencing factors by acquiring global and mechanical / electronic influence information sets; deeply analyzing and revealing the interactions and relationships between these influencing factors by acquiring global and mechanical / electronic influence factor relationship vector sets; and finally, ensuring that the radar antenna array remains in optimal operating condition by acquiring calibration labels and performing dynamic calibration.

[0172] In other, independently achievable embodiments, the step of obtaining the radar antenna array calibration label corresponding to the global influence information set based on the global influence factor relationship vector network and the mechanical and electronic influence factor relationship vector network, through the calibration execution matching sub-algorithm included in the calibration control algorithm, includes: obtaining Q first spatiotemporal signal feature arrays based on the global influence factor relationship vector network and the spatiotemporal feature extraction module included in the calibration control algorithm, wherein each first spatiotemporal signal feature array corresponds to a global influence factor relationship vector network; and obtaining Q first spatiotemporal signal feature arrays based on the global influence factor relationship vector network and the mechanical and electronic influence factor relationship vector network, through the calibration execution matching sub-algorithm included in the calibration control algorithm, through the calibration execution matching sub-algorithm ... mechanical and electronic influence factor relationship vector network, through the calibration execution matching sub-algorithm, through the mechanical and electronic influence factor relationship vector network, through the calibration execution matching sub-algorithm, through the mechanical and electronic influence factor relationship vector network, through the mechanical and electronic influence factor A two-space attention network is used to acquire Q second spatiotemporal signal feature arrays, each of which corresponds to a mechanical and electronic influencing factor relationship vector network. The Q first spatiotemporal signal feature arrays and the Q second spatiotemporal signal feature arrays are integrated to obtain Q target spatiotemporal signal feature arrays, each of which includes one first spatiotemporal signal feature array and one second spatiotemporal signal feature array. Based on the Q target spatiotemporal signal feature arrays, the calibration execution matching sub-algorithm included in the calibration control algorithm is used to obtain the radar antenna array calibration label corresponding to the global influence information set.

[0173] It is understood that this embodiment involves how to obtain the radar antenna array calibration label corresponding to the global influence information set through the calibration execution matching sub-algorithm in the calibration control algorithm, based on the global influence factor relationship vector network set and the mechanical and electronic influence factor relationship vector network set. This process involves the integration of multiple key steps and technologies, aiming to ensure the calibration accuracy and efficiency of the radar antenna array.

[0174] First, the AI ​​correction system, based on a global influencing factor relationship vector network, utilizes the spatiotemporal feature extraction module in the calibration control algorithm to obtain Q first spatiotemporal signal feature arrays. The core of this step lies in the fact that each global influencing factor relationship vector network contains rich spatiotemporal information, which is crucial for understanding the behavior patterns of the radar antenna array globally. The spatiotemporal feature extraction module can deeply mine the spatiotemporal features in these relationship vector networks, transforming them into first spatiotemporal signal feature arrays, providing strong support for subsequent processing.

[0175] Next, the AI ​​calibration system shifts its focus to the set of relational vector networks of electromechanical influencing factors. It then uses the second spatial attention network within the calibration control algorithm to acquire Q second spatiotemporal signal feature arrays. Since the electromechanical influencing factors are closely linked to the physical characteristics and operating status of the radar antenna array, the analysis of their relational vector networks is equally important. The second spatial attention network focuses on capturing the spatial features within these relational vector networks and combining them with temporal features to form the second spatiotemporal signal feature arrays, providing more refined information for the calibration of the radar antenna array.

[0176] After acquiring Q first-space-time signal feature arrays and Q second-space-time signal feature arrays, the AI ​​correction system performs integrated processing to obtain Q target space-time signal feature arrays. The purpose of this step is to fuse the space-time characteristics of global influencing factors and mechanical-electronic influencing factors to form a more comprehensive and accurate representation of space-time signal features. Each target space-time signal feature array contains one first-space-time signal feature array and one second-space-time signal feature array, which together form the basis of radar antenna array calibration.

[0177] Finally, based on Q target spatiotemporal signal feature arrays, the AI ​​calibration system uses the calibration execution matching sub-algorithm in the calibration control algorithm to obtain the radar antenna array calibration labels corresponding to the global influence information set. This step is the core of the entire calibration process, requiring the AI ​​calibration system to transform the spatiotemporal signal features obtained in the previous steps into specific calibration instructions or parameters. The calibration execution matching sub-algorithm deeply analyzes the information in the target spatiotemporal signal feature arrays and matches it with preset calibration standards, ultimately generating calibration labels suitable for the radar antenna array. These calibration labels contain all the necessary calibration information, guiding the radar antenna array to perform precise calibration operations during actual operation.

[0178] Thus, through key steps such as spatiotemporal feature extraction, integrated processing, and calibration execution matching, comprehensive and accurate calibration of the radar antenna array is achieved. This method not only improves the calibration efficiency of radar antenna arrays but also provides strong assurance for their stability and reliability in practical applications.

[0179] In some independent embodiments, the step of obtaining Q first spatiotemporal signal feature arrays based on the global influencing factor relationship vector network set and through the spatiotemporal feature extraction module included in the calibration control algorithm includes: for each global influencing factor relationship vector network set, obtaining a first coded influencing factor relationship vector network through the coding layer included in the spatiotemporal feature extraction module, wherein the spatiotemporal feature extraction module belongs to the calibration control algorithm; for each global influencing factor relationship vector network set, obtaining a first decoded influencing factor relationship vector network through the decoding layer included in the spatiotemporal feature extraction module; for each global influencing factor relationship vector network set, obtaining a first fused influencing factor relationship vector network based on the first coded influencing factor relationship vector network and the first decoded influencing factor relationship vector network through the feature fusion layer included in the spatiotemporal feature extraction module; and for each global influencing factor relationship vector network set, obtaining a first spatiotemporal signal feature array based on the first fused influencing factor relationship vector network and the global influencing factor relationship vector network through the first decoding layer included in the spatiotemporal feature extraction module.

[0180] In practical applications, the technical solution of obtaining Q first spatiotemporal signal feature arrays based on the global influencing factor relationship vector network and through the spatiotemporal feature extraction module in the calibration control algorithm involves the integration of multiple key steps and technologies. It aims to comprehensively and accurately extract the spatiotemporal features in the global influencing factor relationship vector network and provide strong support for the subsequent calibration of radar antenna arrays.

[0181] First, the AI ​​correction system, for each global influencing factor relationship vector network in the global influencing factor relationship vector network set, obtains the first encoded influencing factor relationship vector network through the encoding layer in the spatiotemporal feature extraction module. The encoding layer, a crucial component of the spatiotemporal feature extraction module, performs preliminary processing and transformation on the global influencing factor relationship vector network, encoding key information and features to form the first encoded influencing factor relationship vector network. This step lays the foundation for subsequent decoding and feature fusion.

[0182] Secondly, the AI ​​correction system also targets each global influencing factor relationship vector network in the global influencing factor relationship vector network set, obtaining the first decoded influencing factor relationship vector network through the decoding layer in the spatiotemporal feature extraction module. The decoding layer complements the encoding layer, further decoding and processing the encoded influencing factor relationship vector network to restore its spatiotemporal features and information, forming the first decoded influencing factor relationship vector network. This step provides the necessary decoding information for subsequent feature fusion.

[0183] Then, the AI ​​correction system, for each global influencing factor relationship vector network in the global influencing factor relationship vector network set, obtains a first fused influencing factor relationship vector network based on the first encoded and first decoded influencing factor relationship vector networks through the feature fusion layer in the spatiotemporal feature extraction module. The feature fusion layer is the core part of the spatiotemporal feature extraction module; it effectively fuses the encoded and decoded influencing factor relationship vector networks, extracting key spatiotemporal features and information to form the first fused influencing factor relationship vector network. This step provides crucial feature support for obtaining the first spatiotemporal signal feature array.

[0184] Finally, the AI ​​correction system, still targeting each global influencing factor relationship vector network in the global influencing factor relationship vector network set, obtains the first spatiotemporal signal feature array based on the first fused influencing factor relationship vector network and the global influencing factor relationship vector network through the first decoding layer in the spatiotemporal feature extraction module. The first decoding layer plays a crucial role here, performing the final decoding and processing of the fused influencing factor relationship vector network to extract its spatiotemporal signal features, forming the first spatiotemporal signal feature array. This step is the core output of the entire technical solution, providing important spatiotemporal signal feature support for subsequent radar antenna array calibration.

[0185] Thus, by integrating multiple key steps and technologies such as the encoding layer, decoding layer, feature fusion layer, and first decoding layer, comprehensive and accurate extraction of spatiotemporal features from the global influencing factor relationship vector network set is achieved. This not only improves the accuracy and efficiency of radar antenna array calibration but also provides strong technical support for its stability and reliability in practical applications.

[0186] In summary, in an exemplary application scenario, the AI ​​correction system, through the sensors and measuring devices built into the radar system, acquired the following real-time sensing and measurement status data: pointing angle of 352.6 degrees, signal strength of -45 dBm, phase difference of 2.3 degrees, ambient temperature of 25 degrees Celsius, humidity of 40% RH, and mechanical vibration amplitude of 0.05 mm. This data was organized into a multi-dimensional state attribute vector set, represented as: [352.6, -45, 2.3, 25, 40, 0.05], providing a foundation for subsequent analysis and processing.

[0187] After analyzing the sensor measurement status data, the AI ​​correction system determined the following target performance characteristics: signal transmission efficiency had a feature weight of 0.8, phase consistency had a feature weight of 0.7, and pointing accuracy had a feature weight of 0.9. These feature weights indicate their importance in the overall performance evaluation. For example, pointing accuracy has the highest feature weight, indicating that it has a crucial impact on the performance of X-band radar.

[0188] Next, the AI ​​correction system further analyzed the data and determined the following target correction feature recognition results: the feature weight corresponding to the correction decision feature with signal transmission efficiency was 0.6, the feature weight corresponding to the correction decision feature with phase consistency was 0.5, and the feature weight corresponding to the correction decision feature with pointing accuracy was 0.8. These feature weights reflect the effectiveness and applicability of the corresponding correction decision features.

[0189] After comprehensive evaluation, the AI ​​correction system identified the key X-band performance characteristic as pointing accuracy and determined the desired horizontal correction characteristic as adjusting the pointing angle to reduce deviation.

[0190] The AI ​​calibration system further analyzed the distribution and deviation information of the pointing angle from the sensor measurement status data. It found that the current pointing angle was 352.6 degrees, while the expected pointing angle should be 355 degrees. Therefore, the AI ​​calibration system automatically executed a calibration procedure, adjusting the antenna array's control parameters, such as phase control codes and phase shifter settings, to achieve precise horizontal correction. After calibration, the AI ​​calibration system verified the correction results, confirming that the pointing angle had been successfully adjusted to 355 degrees, and the performance of the antenna array was effectively improved.

[0191] Furthermore, Figure 2 This is a schematic diagram of the structure of an AI correction system 200 provided in an embodiment of this application. Figure 2 The AI ​​correction system 200 shown includes a processor 210, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0192] Optionally, such as Figure 2 As shown, the AI ​​correction system 200 may further include a memory 230. The processor 210 can retrieve and run computer programs from the memory 230 to implement the methods described in this embodiment.

[0193] The memory 230 can be a separate device independent of the processor 210, or it can be integrated into the processor 210.

[0194] Optionally, such as Figure 2 As shown, the AI ​​correction system 200 may also include a transceiver 220, which the processor 210 can control to interact with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices.

[0195] Optionally, the AI ​​correction system 200 can implement the corresponding processes of the storage engine or components (such as processing modules) in the storage engine or the device on which the storage engine is deployed in the various methods of the embodiments of this application. For the sake of brevity, these will not be described in detail here.

[0196] It should be understood that the processor in this application embodiment may be an integrated circuit chip with signal processing capabilities.

[0197] It is understood that the memory in the embodiments of this application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory in the systems and methods described herein is intended to include, but is not limited to, suitable types of memory.

[0198] Based on the above, a readable storage medium is provided, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0199] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0200] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0201] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of protection of this application, and all of these forms are within the protection scope of this application.

Claims

1. A method for intelligent horizontal correction of an X-band radar antenna, characterized in that, The method is applied to an AI correction system, and the method includes: Acquire the sensing measurement state data of the target radar antenna array, and determine the multidimensional state attribute vector set of the sensing measurement state data; The target performance feature identification result is determined based on the multidimensional state attribute vector set, wherein the target performance feature identification result includes several first antenna performance feature units, and the feature weight of each first antenna performance feature unit indicates the confidence that the first antenna performance feature unit is an X-band performance feature. The target calibration feature recognition result is determined based on the multidimensional state attribute vector set, wherein the target calibration feature recognition result includes several second antenna performance feature units, each second antenna performance feature unit corresponds to a first antenna performance feature unit, and the feature weight of each second antenna performance feature unit indicates the calibration decision feature matched by the first antenna performance feature unit corresponding to that second antenna performance feature unit. Based on the target performance feature identification results and the target calibration feature identification results, determine the X-band performance features and the expected horizontal correction features; Based on the X-band performance characteristics and the desired horizontal correction characteristics, the distribution information and deviation information of the data to be adjusted are determined in the sensing measurement state data, and the target radar antenna array is calibrated based on the distribution information and the deviation information; The target performance feature identification result determined based on the multidimensional state attribute vector set includes: A feature embedding operation is performed on the multidimensional state attribute vector set to determine the multidimensional state attribute vector set with completed feature embedding, wherein the multidimensional state attribute vector set with completed feature embedding includes several third antenna performance feature units. For each third antenna performance feature unit in the multidimensional state attribute vector set with completed feature embedding, the target pointing angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the pointing dimension is determined, the target phase angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the phase dimension is determined, and the feature weight of the first antenna performance feature unit in the target performance feature recognition result is determined based on the target pointing angle feature variable and the target phase angle feature variable. Specifically, determining the target pointing angle feature variable in the pointing dimension of the multidimensional state attribute vector set after feature embedding includes: traversing the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first pointing feature label of the multidimensional state attribute vector set based on the pointing dimension, and determining the maximum value among the traversed feature weights as the first target pointing angle feature variable. And / or, based on the pointing dimension, traverse the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second pointing feature label of the multidimensional state attribute vector set, and determine the maximum value among the traversed feature weights as the second target pointing angle feature variable. Determining the target phase angle feature variable of the multidimensional state attribute vector set with completed feature embedding in the phase dimension includes: based on the phase dimension, traversing the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the first phase feature label of the multidimensional state attribute vector set, and determining the maximum value among the traversed feature weights as the first target phase angle feature variable. And / or based on the phase dimension, traverse the feature weights of all third antenna performance feature units between the third antenna performance feature unit and the second phase feature label of the multidimensional state attribute vector set, and determine the maximum value among the traversed feature weights as the second target phase angle feature variable. Determining the feature weights of the first antenna performance feature unit in the target performance feature identification result based on the target pointing angle feature variable and the target phase angle feature variable includes: weighting the first target pointing angle feature variable and / or the second target pointing angle feature variable with the first target phase angle feature variable and / or the second target phase angle feature variable to determine the feature weights of the first antenna performance feature unit corresponding to the third antenna performance feature unit in the target performance feature identification result. The multidimensional state attribute vector set used to determine the sensing measurement state data includes: The sensing measurement state data is processed by at least one feature embedding branch to obtain sensing measurement state data with completed feature embedding. The sensor measurement state data with completed feature embedding is subjected to at least one feature pooling and at least one feature derivation through at least one feature pooling branch and at least one feature derivation branch to determine a multidimensional state attribute vector set of the sensor measurement state data, wherein the feature granularity of the multidimensional state attribute vector set is smaller than that of the sensor measurement state data. The multidimensional state attribute vector set, the target performance feature recognition result, and the target calibration feature recognition result are generated through at least one long short-term memory model, which is debugged through the following steps: Determine a set of debugging samples, wherein the set of debugging samples includes at least one sample of sensor measurement state data, and the at least one sample of sensor measurement state data carries prior distribution information and prior deviation information of the prior data to be adjusted; For each sample of sensor measurement state data: Determine the multidimensional state attribute vector set sample of the sensor measurement state data sample; Based on the multidimensional state attribute vector set examples, determine the target performance feature identification result examples; Based on the multidimensional state attribute vector set examples, determine the target proofreading feature recognition result examples; Based on the target performance feature identification result examples and the target calibration feature identification result examples, determine the X-band performance feature examples and the horizontal correction expected feature examples; Based on the X-band performance characteristic sample and the horizontal correction expected characteristic sample, the distribution information sample and deviation information sample of the data sample to be adjusted are determined in the sensing measurement state data sample; Optimize the model weights of the at least one long short-term memory model so that the differences between the distribution information samples and deviation information samples of the data samples to be adjusted in the sensing measurement state data samples and the prior distribution information and prior deviation information of the prior data to be adjusted converge. The difference includes at least one of the following comparison errors: the distribution comparison error between the distribution information sample and the prior distribution information; and the deviation comparison error between the deviation information sample and the prior deviation information.

2. The method as described in claim 1, characterized in that, The target performance feature identification result includes at least one performance attention index, and the feature weight of the first antenna performance feature unit in each performance attention index indicates the confidence level that the first antenna performance feature unit is the X-band performance feature of the target of the corresponding performance type.

3. The method as described in claim 1, characterized in that, The target proofreading feature recognition result determined based on the multidimensional state attribute vector set includes: The multidimensional state attribute vector set is subjected to feature embedding operation to obtain the target calibration feature recognition result. For each performance type of X-band performance feature, the target calibration feature recognition result includes at least one performance attention index. The feature weight of the second antenna performance feature unit of each performance attention index indicates that the first antenna performance feature unit corresponding to the second antenna performance feature unit serves as the performance correction reference under the set correction rule matched by the X-band performance feature of that performance type.

4. The method as described in claim 1, characterized in that, Determining the X-band performance characteristics and horizontal correction expectation characteristics based on the target performance characteristic identification results and the target correction characteristic identification results includes: Based on the feature weight of each first antenna performance feature unit in the target performance feature identification result, the discrimination score of the first antenna performance feature unit belonging to the X-band performance feature is determined; The first antenna performance feature element with a discrimination score greater than the set discrimination score is identified as the X-band performance feature; and based on the distribution characteristics of the X-band performance feature in the target performance feature identification result, the horizontal correction expectation feature corresponding to the X-band performance feature is determined in the target calibration feature identification result.

5. The method as described in claim 1, characterized in that, Based on the X-band performance characteristics and the desired horizontal correction characteristics, the distribution information and deviation information of the data to be adjusted in the sensing measurement state data are determined, including: Determine the feature association description between the multidimensional state attribute vector set and the sensing measurement state data; Based on the X-band performance characteristics, the expected horizontal correction characteristics, and the feature correlation description, the distribution information and initial noise information of the data to be adjusted in the sensing measurement state data are determined; Based on the multidimensional state attribute vector set, a noise perturbation for the initial noise information is determined; Based on the noise disturbance and the initial noise information, the deviation information of the data to be adjusted is determined in the sensing measurement state data.

6. The method as described in claim 1, characterized in that, The X-band performance characteristics and the horizontal correction expectation characteristics include X-band performance characteristics for the data to be adjusted for at least one time-series label and horizontal correction expectation characteristics for the data to be adjusted for that performance category.

7. The method as described in claim 1, characterized in that, The sensing measurement state data is radar antenna array state data collected by the built-in measurement device. Before determining the multidimensional state attribute vector set of the sensing measurement state data, the method further includes: adjusting according to the original feature value of each antenna performance feature unit in the sensing measurement state data.

8. An AI correction system, characterized in that, The method includes at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, causing the at least one processor to perform the method according to any one of claims 1-7.

Images