Microwave antenna beam control method and system based on machine learning

By screening and normalizing the characteristics, using neural network models to perform microwave antenna beam regulation, solving the problem of low regulation accuracy in the existing technology and achieving a more efficient beam regulation effect.

CN117972379BActive Publication Date: 2024-11-22NANTONG RES INST FOR ADVANCED COMM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311697424.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-11
Publication Date
2024-11-22
Estimated Expiration
2043-12-11

AI Technical Summary

Technical Problem

In the prior art, the machine learning microwave antenna beam regulation model requires too much data and the characteristics have not been screened, resulting in low regulation accuracy, poor adaptability and poor regulation effect.

Method used

By obtaining the data required for modeling, feature extraction and classification, filtering effective features, establishing new features, and normalizing them, dividing the features into training set, verification set and test set, and using neural network models to perform microwave antenna beams Regulation, effective features include signal strength, signal-to-noise ratio and multipath propagation characteristics.

Benefits of technology

Improve the accuracy and adaptability of beam regulation, ensuring the improvement of regulation effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117972379B_ABST
    Figure CN117972379B_ABST
Patent Text Reader

Abstract

The present invention discloses a microwave antenna beam control method and system based on machine learning, which relates to the field of data processing technology, including obtaining data required for modeling, extracting features according to the data required for modeling; classifying the extracted features according to functional categories, screening all features in each category, and retaining effective features; establishing new features according to the association between effective features in multiple categories, and performing normalized scale processing based on new features and effective features; dividing new features and effective features into training sets, verification sets, and test sets; training a neural network model according to the training sets, verification sets, and test sets, and performing microwave antenna beam control through the neural network model. The accuracy and adaptability of beam control are improved, and the control effect is guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and more specifically, to a microwave antenna beam control method and system based on machine learning. Background Art

[0002] Microwave antenna beam steering based on machine learning is a cutting-edge technology that combines knowledge from the fields of microwave communications, antenna technology, and machine learning. It enables smarter, adaptive, and high-performance communication and radar systems by adjusting the antenna radiation pattern. This technology requires a deep understanding of microwave communications, beam steering principles, and machine learning and deep learning methods, while relying on environmental perception, big data, and computing power. It has broad application prospects in the fields of communications, military, aerospace, etc., and can improve system performance and flexibility.

[0003] In the prior art, the machine learning control model requires too much data, and the features are not screened, resulting in the entry of irrelevant features, which makes the control model have low control accuracy, poor adaptability, and poor control effect.

[0004] Therefore, how to improve the control accuracy is a technical problem that needs to be solved. Summary of the invention

[0005] The present invention provides a microwave antenna beam control method based on machine learning to solve the technical problem of low beam control accuracy in the prior art. The method comprises:

[0006] Obtain the data required for modeling, and perform feature extraction based on the data required for modeling;

[0007] Classify the extracted features according to functional categories, filter all features in each category, and retain the valid features;

[0008] New features are established based on the association between valid features in multiple categories, and normalized scale processing is performed based on the new features and valid features;

[0009] Divide new features and valid features into training sets, validation sets, and test sets;

[0010] Train the neural network model based on the training set, validation set and test set, and perform microwave antenna beam control through the neural network model;

[0011] Among them, effective features include signal strength features, signal-to-noise ratio features, and multipath propagation features.

[0012] In some embodiments of the present application, all features in each category are screened and valid features are retained, including:

[0013] Calculate the correlation between each feature and the beam steering index, and record the correlation exceeding the first threshold as the first correlation;

[0014] The correlation degree exceeding the second threshold and not exceeding the first threshold is recorded as the second correlation degree, and the correlation degree not exceeding the second threshold is recorded as the third correlation degree;

[0015] The effective feature is determined according to the first correlation, the second correlation and the third correlation.

[0016] In some embodiments of the present application, determining effective features according to the first relevance, the second relevance, and the third relevance includes:

[0017] Determine a first correlation average value according to all first correlations, determine a second correlation average value according to all second correlations, and determine a third correlation average value according to all third correlations;

[0018] Determine a first ratio based on a difference between a first correlation average and a second correlation average, determine a second ratio based on a difference between a second correlation average and a third correlation average, and determine a third ratio based on a difference between the first correlation average and the third correlation average;

[0019] Sorting according to the magnitude of the respective correlations among the first correlation, the second correlation and the third correlation;

[0020] According to the first ratio, the features corresponding to the top multiple first correlations are used as effective features;

[0021] According to the second ratio, the features corresponding to the top multiple second correlations are used as effective features;

[0022] According to the third ratio, the features corresponding to the top multiple third correlations are taken as valid features.

[0023] In some embodiments of the present application, before screening all features in each category, the method further includes:

[0024] Obtain all beam steering effect indicators, and determine the beam steering indicator according to all beam steering effect indicators;

[0025] ;

[0026] Among them, P is the beam steering index, n is the number of all beam steering effect indicators, is the original weight corresponding to the i-th beam steering effect index, exp is the exponential function, is the beam steering effect index parameter of the i-th beam that exceeds the corresponding threshold, is the threshold corresponding to the i-th beam control effect index, k is a preset constant, is the i-th beam steering effect indicator, which includes main lobe gain, side lobe level, beam width, and beam direction error.

[0027] In some embodiments of the present application, new features are established based on the association between valid features in multiple categories, including:

[0028] The correlations between effective features of different categories are calculated respectively, and those with correlations exceeding the corresponding threshold are used as new features.

[0029] In some embodiments of the present application, new features and valid features are divided into training sets, validation sets and test sets, including:

[0030] Divide new features and valid features according to the standard ratio to obtain training set, validation set and test set;

[0031] Determine the performance of the initial model in the training set, validation set, and test set, respectively, and record them as the first performance, second performance, and third performance;

[0032] If the first performance meets the first requirement, and the second performance and the third performance do not meet the second requirement and the third requirement, then an adjustment coefficient is determined according to the difference between the first performance and the second performance and the difference between the second performance and the third performance, and the standard ratio is adjusted based on the adjustment coefficient, so as to re-divide the training set, the validation set and the test set;

[0033] If the first performance does not meet the first requirement, and the second and third performances do not meet the second and third requirements, the standard ratio is adjusted according to the first, second and third performances, thereby re-dividing the training set, validation set and test set.

[0034] In some embodiments of the present application, the method further includes:

[0035] If the first performance meets the first requirement, and the second and third performances meet the second and third requirements, the standard ratio is not adjusted.

[0036] Correspondingly, the present application also provides a microwave antenna beam steering system based on machine learning, the system comprising:

[0037] The first module is used to obtain the data required for modeling and perform feature extraction based on the data required for modeling;

[0038] The second module is used to classify the extracted features according to functional categories, filter all features in each category, and retain the valid features;

[0039] The third module is used to establish new features according to the association between valid features in multiple categories, and perform normalized scale processing based on the new features and the valid features;

[0040] The fourth module is used to divide new features and valid features into training sets, validation sets and test sets;

[0041] The fifth module is used to train the neural network model according to the training set, the validation set and the test set, and perform microwave antenna beam control through the neural network model;

[0042] Among them, effective features include signal strength features, signal-to-noise ratio features, and multipath propagation features.

[0043] By applying the above technical solutions, the data required for modeling is obtained, and features are extracted based on the data required for modeling; the extracted features are classified according to functional categories, all features in each category are screened, and effective features are retained; new features are established based on the association between effective features in multiple categories, and normalized scale processing is performed based on new features and effective features; new features and effective features are divided into training sets, verification sets, and test sets; neural network models are trained based on training sets, verification sets, and test sets, and microwave antenna beam control is performed through the neural network model. The accuracy and adaptability of beam control are improved, and the control effect is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A schematic diagram of a process flow of a microwave antenna beam control method based on machine learning proposed in an embodiment of the present invention is shown;

[0046] Figure 2 A schematic structural diagram of a microwave antenna beam control system based on machine learning proposed in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] The present application embodiment provides a microwave antenna beam control method based on machine learning, such as Figure 1 As shown, the method comprises the following steps:

[0049] Step S101, obtaining data required for modeling, and performing feature extraction based on the data required for modeling.

[0050] In this embodiment, the data required for modeling includes antennas, communication environment, signal characteristics, beam steering algorithms, etc. The following are some of the main data factors:

[0051] Antenna design and performance:

[0052] The type of antenna, the number of antenna elements, and the arrangement of the antenna directly affect the ability to steer the beam. Antennas with high gain and strong directivity usually have better beam steering effects.

[0053] The operating frequency range and bandwidth of the antenna can also affect performance, as beam characteristics may differ at different frequencies.

[0054] Communication environment:

[0055] Obstacles, reflections, diffraction, and multipath propagation in the communication environment will affect the path and characteristics of signal propagation, thereby affecting the effect of beam steering.

[0056] Weather conditions, such as rain, snow, atmospheric humidity, etc., may also cause signal attenuation and changes in propagation performance.

[0057] Signal characteristics:

[0058] The signal's frequency, bandwidth, modulation method and other characteristics will affect the demand and effect of beam steering.

[0059] Different signal parameters such as power and modulation depth may also require different beam settings.

[0060] Beam steering algorithm:

[0061] The beam steering algorithm and strategy chosen directly determine the shape, direction, and width of the beam.

[0062] Adaptive beam steering algorithms can automatically adjust beams to adapt to changes in dynamic environments, but their performance also depends on the design of the algorithm and parameter settings.

[0063] System hardware performance:

[0064] The performance of the antenna array hardware, including the sensitivity, dynamic range, and phase stability of the antenna elements, has a significant impact on the effectiveness of beam steering.

[0065] The performance of the power amplifier and transceiver will also affect the signal quality and transmission efficiency.

[0066] Signal Interference and Noise:

[0067] Signal interference sources (such as other communication systems or electronic devices) as well as noise in the environment can degrade signal quality and thus affect the performance of beam steering.

[0068] Step S102, classify the extracted features according to functional categories, filter all features in each category, and retain valid features.

[0069] In this embodiment, effective features include the following possible feature extraction examples (including but not limited to):

[0070] Signal strength characteristics:

[0071] Average Signal Strength: Calculates the average signal strength.

[0072] Maximum Signal Strength: Determines the maximum signal strength in the data.

[0073] Minimum Signal Strength: Determines the minimum signal strength in the data.

[0074] Signal strength variation: describes the changing trend or variance of signal strength.

[0075] Signal-to-noise ratio characteristics:

[0076] Average SNR: Calculates the average of the signal-to-noise ratio.

[0077] Maximum Signal-to-Noise Ratio: Determines the maximum signal-to-noise ratio in the data.

[0078] Minimum Signal-to-Noise Ratio: Determines the minimum signal-to-noise ratio in the data.

[0079] Signal-to-noise ratio variation: describes the variation trend or variance of the signal-to-noise ratio.

[0080] Multipath propagation characteristics:

[0081] Multipath Delay: Identifies the delay time in multipath propagation.

[0082] Multipath Amplitude: Identifies the signal amplitudes of different paths in multipath propagation.

[0083] Multipath Phase: Identifies the signal phases of different paths in multipath propagation.

[0084] Weather Conditions Characteristics:

[0085] Temperature: Measures temperature information in weather.

[0086] Humidity: Measures humidity information in the weather.

[0087] Wind Speed ​​and Direction: Measures wind speed and direction.

[0088] Barometric Pressure: Measures the pressure of the atmosphere.

[0089] Time characteristics:

[0090] Timestamp: Associate data with time to account for the effects of time.

[0091] Seasonal characteristics: such as season, month, day of the week, etc., are used to take into account seasonal changes.

[0092] Space features:

[0093] Location coordinates: If your data includes location information, you can use coordinates such as longitude and latitude as features.

[0094] Distance: Distance information from a specific location or device.

[0095] Frequency characteristics:

[0096] Frequency analysis: Analyze the frequency distribution of the signal and extract frequency domain features, such as spectrum density.

[0097] In some embodiments of the present application, all features in each category are screened and valid features are retained, including:

[0098] Calculate the correlation between each feature and the beam steering index, and record the correlation exceeding the first threshold as the first correlation;

[0099] The correlation degree exceeding the second threshold and not exceeding the first threshold is recorded as the second correlation degree, and the correlation degree not exceeding the second threshold is recorded as the third correlation degree;

[0100] The effective feature is determined according to the first correlation, the second correlation and the third correlation.

[0101] In some embodiments of the present application, determining effective features according to the first relevance, the second relevance, and the third relevance includes:

[0102] Determine a first correlation average value according to all first correlations, determine a second correlation average value according to all second correlations, and determine a third correlation average value according to all third correlations;

[0103] Determine a first ratio based on a difference between a first correlation average and a second correlation average, determine a second ratio based on a difference between a second correlation average and a third correlation average, and determine a third ratio based on a difference between the first correlation average and the third correlation average;

[0104] Sorting according to the magnitude of the respective correlations among the first correlation, the second correlation and the third correlation;

[0105] According to the first ratio, the features corresponding to the top multiple first correlations are used as effective features;

[0106] According to the second ratio, the features corresponding to the top multiple second correlations are used as effective features;

[0107] According to the third ratio, the features corresponding to the top multiple third correlations are taken as valid features.

[0108] In this embodiment, each difference value corresponds to a ratio in the corresponding relationship.

[0109] In some embodiments of the present application, before screening all features in each category, the method further includes:

[0110] Obtain all beam steering effect indicators, and determine the beam steering indicator according to all beam steering effect indicators;

[0111] ;

[0112] Among them, P is the beam steering index, n is the number of all beam steering effect indicators, is the original weight corresponding to the i-th beam steering effect index, exp is the exponential function, is the beam steering effect index parameter of the i-th beam that exceeds the corresponding threshold, is the threshold corresponding to the i-th beam control effect index, k is a preset constant, is the i-th beam control effect index.

[0113] In this embodiment, Indicates the adjustment of the weight of an overly large effect indicator.

[0114] In this embodiment, all beam steering effect indicators include (including but not limited to):

[0115] The effectiveness of microwave antenna beam steering can be evaluated by multiple indicators, which can be used to measure the performance of antenna systems in different applications. The following are some common indicators used to evaluate the effectiveness of microwave antenna beam steering:

[0116] Main Beam Gain: Main lobe gain refers to the gain of the radiated power of the beam in the main lobe direction relative to the omnidirectional radiation. An increase in the main lobe gain indicates that the beam is successfully focused in the desired direction.

[0117] Sidelobe Level: Sidelobe is the radiation direction outside the main lobe, and the sidelobe level is a parameter that describes the intensity of sidelobe radiation. A lower sidelobe level means smaller sidelobe radiation, which helps reduce interference.

[0118] Beamwidth: Beamwidth indicates the angular range of the main lobe and is usually measured as the half power beamwidth (HPBW) of the main lobe. A smaller beamwidth indicates a more directional beam.

[0119] Beam Steering Error: Beam steering error indicates the deviation between the actual beam direction and the expected direction. A smaller beam steering error indicates a better beam steering effect.

[0120] Beam Tracking Rate: For applications that require tracking of moving targets, the beam tracking rate indicates how fast the system adjusts the beam direction. A higher tracking rate indicates better target tracking capabilities.

[0121] Adaptation Performance: The performance of adaptive beam steering can be evaluated by multiple indicators, including error rate, signal-to-noise ratio improvement, interference suppression, etc.

[0122] Beam Switching Time: If the system supports switching between different beams, the beam switching time indicates the time required to switch from one beam to another. A shorter switching time helps to quickly adapt to different situations.

[0123] Signal Quality Metrics: Some specific applications may use specific signal quality metrics, such as bit error rate, bit error rate, communication link quality, etc. to evaluate the beam steering effect.

[0124] Communication Performance Metrics: For communication systems, communication performance metrics such as transmission rate, connectivity, data throughput, etc. can be used to evaluate the effectiveness of beam steering.

[0125] Radar Performance Metrics: For radar systems, performance metrics such as target detection probability, false alarm rate, and range resolution are used to evaluate the effectiveness of beam steering in target detection and tracking.

[0126] Step S103: establishing new features according to the associations between the valid features in the multiple categories, and performing normalization and scaling processing based on the new features and the valid features.

[0127] In this embodiment, the new features, for example, create interaction features between different features, such as interaction features between signal strength and weather conditions, to capture the correlation between features.

[0128] Normalization is an important data preprocessing technique that is used to scale different features or attributes of data to similar scales so that they can be better used for training and analysis of machine learning models. Normalization helps to prevent the model from being affected by differences in the scales of different features and improves the stability and performance of the model. It includes the following methods: Min-Max Scaling, Z-Score Standardization, Robust Scaling, etc.

[0129] Step S104, dividing the new features and valid features into a training set, a validation set and a test set.

[0130] In some embodiments of the present application, new features are established based on the association between valid features in multiple categories, including:

[0131] The correlations between effective features of different categories are calculated respectively, and those with correlations exceeding the corresponding threshold are used as new features.

[0132] In some embodiments of the present application, new features and valid features are divided into training sets, validation sets and test sets, including:

[0133] Divide new features and valid features according to the standard ratio to obtain training set, validation set and test set;

[0134] Determine the performance of the initial model in the training set, validation set, and test set, respectively, and record them as the first performance, second performance, and third performance;

[0135] If the first performance meets the first requirement, and the second performance and the third performance do not meet the second requirement and the third requirement, then an adjustment coefficient is determined according to the difference between the first performance and the second performance and the difference between the second performance and the third performance, and the standard ratio is adjusted based on the adjustment coefficient, so as to re-divide the training set, the validation set and the test set;

[0136] If the first performance does not meet the first requirement, and the second and third performances do not meet the second and third requirements, the standard ratio is adjusted according to the first, second and third performances, thereby re-dividing the training set, validation set and test set.

[0137] In this embodiment, the difference between the first performance and the second performance, and the difference between the second performance and the third performance determine the adjustment coefficient, and the two differences together correspond to one adjustment coefficient.

[0138] In this embodiment, this is to avoid overfitting or underfitting.

[0139] 1. Overfitting:

[0140] Impact on the training set: Overfitting can cause the model to perform well on the training set and fit the training data very accurately, but may capture noise or anomalies in the data and introduce too much complexity. The performance of the model on the training set is high.

[0141] Impact on validation set: An overfit model usually performs poorly on the validation set because the model has overfitted to the training data and cannot generalize well to new data. The performance metrics on the validation set may be low.

[0142] Impact on the test set: An overfit model will usually perform poorly on the test set as well, since it has limited generalization capabilities and cannot adapt to data that is different from the training set. Performance metrics on the test set may also be lower.

[0143] 2. Underfitting:

[0144] Impact on the training set: Underfitting means that the model cannot fit the training data well, usually because the model is too simple or not complex enough. Therefore, the performance on the training set may not be very good, and the training error is high.

[0145] Impact on validation set: On the validation set, the performance of the underfit model may still be poor because it cannot capture the complexity and key features of the data. The performance indicators on the validation set may be lower.

[0146] Impact on the test set: An underfit model usually does not perform well on the test set because its generalization ability is limited and it cannot adapt to new data. The performance indicators on the test set may be lower.

[0147] In some embodiments of the present application, the method further includes:

[0148] If the first performance meets the first requirement, and the second and third performances meet the second and third requirements, the standard ratio is not adjusted.

[0149] Step S105, training a neural network model according to the training set, the validation set and the test set, and performing microwave antenna beam control through the neural network model.

[0150] In this embodiment, during operation, the system collects real-time data, such as signal strength, interference, etc. A machine learning model is used to adjust beam parameters based on real-time data to optimize performance. A feedback mechanism can be used to continuously optimize the performance of the model.

[0151] In this embodiment, the beam parameters include multiple parameters involved in microwave antenna beam control, which are used to control the radiation pattern of the antenna to meet the needs of a specific communication, radar or wireless communication system. The following are some common antenna beam parameters:

[0152] Beam Direction: Beam direction refers to the direction of the antenna's main lobe (primary radiation direction), usually described by directional angles, such as azimuth and elevation. The direction of the beam determines the main transmission or reception direction of the signal.

[0153] Beamwidth: Beamwidth is the angular extent of the main lobe, usually measured as the half power width (HPBW) of the main lobe. A smaller beamwidth indicates a more directional beam, while a larger width indicates a wider beam.

[0154] Main Beam Gain: Main lobe gain refers to the gain of the radiated power of the beam in the main lobe direction relative to the omnidirectional radiation. It determines the reception or transmission performance of the antenna in the main lobe direction.

[0155] Sidelobe Level: Sidelobe is the radiation direction outside the main lobe, usually the direction where the signal is not expected to radiate. Sidelobe level is a parameter that describes the intensity of sidelobe radiation, usually expressed in dB. A lower sidelobe level indicates a smaller sidelobe radiation.

[0156] Beam Shape: Beam shape refers to the distribution characteristics of the beam in the horizontal and vertical directions, such as Gaussian beam, uniform beam, etc. Different beam shapes are suitable for different application requirements.

[0157] Beam Steering: Beam steering is the process of adjusting the antenna beam direction to accommodate different communication targets or to track moving targets. This usually involves adjusting the phase difference of the antenna elements.

[0158] Multiple Beams: Some systems support forming multiple beams simultaneously to handle multiple signal sources or perform multiple tasks. Multibeam parameters include the direction and width of each beam.

[0159] Adaptive Beamforming: Adaptive beamforming uses a feedback mechanism to adjust beam parameters in real time to adapt to changing signal, interference and environmental conditions. Adaptive parameters may include phase, amplitude, beam direction, etc.

[0160] Beam Tracking: In some applications, such as radar tracking a moving target, beam tracking is a key parameter. It refers to the speed at which the beam direction is adjusted to track the target's motion.

[0161] Beam Switching: Some systems can switch between different beams to meet different communication needs or coverage areas.

[0162] By applying the above technical solutions, the data required for modeling is obtained, and features are extracted based on the data required for modeling; the extracted features are classified according to functional categories, all features in each category are screened, and effective features are retained; new features are established based on the association between effective features in multiple categories, and normalized scale processing is performed based on new features and effective features; new features and effective features are divided into training sets, verification sets, and test sets; neural network models are trained based on training sets, verification sets, and test sets, and microwave antenna beam control is performed through the neural network model. The accuracy and adaptability of beam control are improved, and the control effect is guaranteed.

[0163] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0164] In order to further explain the technical idea of ​​the present invention, the technical solution of the present invention is now described in combination with specific application scenarios.

[0165] Correspondingly, the present application also provides a microwave antenna beam control system based on machine learning, such as Figure 2 As shown, the system comprises:

[0166] The first module 201 is used to obtain data required for modeling and perform feature extraction based on the data required for modeling;

[0167] The second module 202 is used to classify the extracted features according to functional categories, filter all features in each category, and retain valid features;

[0168] The third module 203 is used to establish new features according to the association between the valid features in multiple categories, and perform normalization scale processing based on the new features and the valid features;

[0169] The fourth module 204 is used to divide the new features and valid features into a training set, a validation set and a test set;

[0170] The fifth module 205 is used to train the neural network model according to the training set, the validation set and the test set, and perform microwave antenna beam control through the neural network model.

[0171] Those skilled in the art will appreciate that the modules in the system in the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A microwave antenna beam control method based on machine learning, characterized in that: The method comprises: Obtain the data required for modeling, and perform feature extraction based on the data required for modeling; Classify the extracted features according to functional categories, filter all features in each category, and retain the valid features; New features are established based on the association between valid features in multiple categories, and normalized scale processing is performed based on the new features and valid features; Divide new features and valid features into training sets, validation sets, and test sets; Train the neural network model based on the training set, validation set and test set, and perform microwave antenna beam control through the neural network model; Among them, effective features include signal strength features, signal-to-noise ratio features, and multipath propagation features; Filter all features within each category and retain valid features, including: Calculate the correlation between each feature and the beam steering index, and record the correlation exceeding the first threshold as the first correlation; The correlation degree exceeding the second threshold and not exceeding the first threshold is recorded as the second correlation degree, and the correlation degree not exceeding the second threshold is recorded as the third correlation degree; Determining effective features according to the first correlation, the second correlation, and the third correlation; Determining effective features according to the first relevance, the second relevance, and the third relevance includes: Determine a first correlation average value according to all first correlations, determine a second correlation average value according to all second correlations, and determine a third correlation average value according to all third correlations; Determine a first ratio based on a difference between a first correlation average and a second correlation average, determine a second ratio based on a difference between a second correlation average and a third correlation average, and determine a third ratio based on a difference between the first correlation average and the third correlation average; Sorting according to the magnitude of the respective correlations among the first correlation, the second correlation and the third correlation; According to the first ratio, the features corresponding to the top multiple first correlations are used as effective features; According to the second ratio, the features corresponding to the top multiple second correlations are used as effective features; According to the third ratio, the features corresponding to the top multiple third relevances are taken as valid features; Before screening all features within each category, the method further includes: Obtain all beam steering effect indicators, and determine the beam steering indicator according to all beam steering effect indicators; ; Among them, P is the beam steering index, n is the number of all beam steering effect indicators, is the original weight corresponding to the i-th beam steering effect index, exp is the exponential function, is the beam steering effect index parameter of the i-th beam that exceeds the corresponding threshold, is the threshold corresponding to the i-th beam control effect index, k is a preset constant, is the i-th beam steering effect indicator, which includes main lobe gain, side lobe level, beam width, and beam direction error.

2. The microwave antenna beam control method based on machine learning according to claim 1, characterized in that: New features are created based on the associations between valid features in multiple categories, including: The correlations between effective features of different categories are calculated respectively, and those with correlations exceeding the corresponding threshold are used as new features.

3. The microwave antenna beam control method based on machine learning according to claim 1, characterized in that: Divide new features and valid features into training sets, validation sets, and test sets, including: Divide new features and valid features according to the standard ratio to obtain training set, validation set and test set; Determine the performance of the initial model in the training set, validation set, and test set, respectively, and record them as the first performance, second performance, and third performance; If the first performance meets the first requirement, and the second performance and the third performance do not meet the second requirement and the third requirement, then an adjustment coefficient is determined according to the difference between the first performance and the second performance and the difference between the second performance and the third performance, and the standard ratio is adjusted based on the adjustment coefficient, so as to re-divide the training set, the validation set and the test set; If the first performance does not meet the first requirement, and the second and third performances do not meet the second and third requirements, the standard ratio is adjusted according to the first, second and third performances, thereby re-dividing the training set, validation set and test set.

4. The method for microwave antenna beam control based on machine learning according to claim 3, characterized in that: The method further comprises: If the first performance meets the first requirement, and the second and third performances meet the second and third requirements, the standard ratio is not adjusted.

5. A microwave antenna beam steering system based on machine learning, characterized in that: The system comprises: The first module is used to obtain the data required for modeling and perform feature extraction based on the data required for modeling; The second module is used to classify the extracted features according to functional categories, filter all features in each category, and retain the valid features; Filter all features within each category and retain valid features, including: Calculate the correlation between each feature and the beam steering index, and record the correlation exceeding the first threshold as the first correlation; The correlation degree exceeding the second threshold and not exceeding the first threshold is recorded as the second correlation degree, and the correlation degree not exceeding the second threshold is recorded as the third correlation degree; Determining effective features according to the first correlation, the second correlation, and the third correlation; Determining effective features according to the first relevance, the second relevance, and the third relevance includes: Determine a first correlation average value according to all first correlations, determine a second correlation average value according to all second correlations, and determine a third correlation average value according to all third correlations; Determine a first ratio based on a difference between a first correlation average and a second correlation average, determine a second ratio based on a difference between a second correlation average and a third correlation average, and determine a third ratio based on a difference between the first correlation average and the third correlation average; Sorting according to the magnitude of the respective correlations among the first correlation, the second correlation and the third correlation; According to the first ratio, the features corresponding to the top multiple first correlations are used as effective features; According to the second ratio, the features corresponding to the top multiple second correlations are used as effective features; According to the third ratio, the features corresponding to the top multiple third relevances are taken as valid features; Before filtering all features within each category, also include: Obtain all beam steering effect indicators, and determine the beam steering indicator according to all beam steering effect indicators; ; Among them, P is the beam steering index, n is the number of all beam steering effect indicators, is the original weight corresponding to the i-th beam steering effect index, exp is the exponential function, is the beam steering effect index parameter of the i-th beam that exceeds the corresponding threshold, is the threshold corresponding to the i-th beam control effect index, k is a preset constant, is the i-th beam steering effect index, which includes main lobe gain, side lobe level, beam width, and beam direction error; The third module is used to establish new features according to the association between valid features in multiple categories, and perform normalized scale processing based on the new features and the valid features; The fourth module is used to divide new features and valid features into training sets, validation sets and test sets; The fifth module is used to train the neural network model according to the training set, the validation set and the test set, and perform microwave antenna beam control through the neural network model; Among them, effective features include signal strength features, signal-to-noise ratio features, and multipath propagation features.