Method for collecting and processing meteorological data of low-altitude aircraft routes

Abstract

The present invention discloses a method for collecting and processing meteorological data for low-altitude aircraft routes, which relates to the technical field of route meteorological data collection and includes the following steps: obtaining carrier phase observations of target satellites from a GNSS receiver, recording corresponding timestamps, and performing original turnover ambiguity resolution on the obtained carrier phase observations; performing first-order difference processing on the carrier phase observations of the same satellite at two consecutive time points to construct a sequence of instantaneous carrier phase variation that reflects the phase change rate. The present invention realizes dynamic weight adjustment of multipath interference satellites through carrier phase disturbance feature extraction and deep learning multipath risk prediction, improves the system's anti-interference capability and data processing robustness, significantly reduces the risk of spatial mismatch and meteorological misjudgment, and ensures positioning accuracy and track perception reliability in complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of airway meteorological data collection, and in particular to a method for collecting and processing airway meteorological data of low-altitude aircraft. Background Art

[0002] The collection and processing of meteorological data along the routes of low-altitude aircraft refers to the real-time monitoring of meteorological conditions along the flight routes by installing various meteorological collection devices on low-altitude aircraft. These devices typically include temperature sensors, humidity sensors, anemometers, barometers, etc., which can collect data on various meteorological parameters such as wind speed, temperature, humidity, and air pressure. The collected data needs to be processed and analyzed to ensure that the aircraft can make appropriate course adjustments based on the current meteorological conditions and optimize flight safety and efficiency. The core purpose of the collection and processing of meteorological data along the routes of low-altitude aircraft is to provide meteorological information support for the aircraft's flight route in order to predict possible meteorological changes and avoid the impact of adverse weather on the aircraft during flight. Through data processing, the original collected data can be converted into easy-to-understand meteorological status information for pilots or automated systems to make decisions.

[0003] The existing technology has the following deficiencies:

[0004] During the collection of en-route meteorological data for low-altitude aircraft, satellite-based GPS navigation systems are susceptible to interference from signal multipath effects in areas with complex terrain (such as urban canyons with tall buildings, cliff edges with sudden changes in terrain, or near large metal structures). This effect occurs when positioning signals from navigation satellites are reflected by surrounding buildings or terrain before reaching the receiver. This makes it impossible for the receiving device to accurately distinguish the path differences between the direct signal and the reflected signal, which in turn causes positioning errors and spatial registration offsets. Because the collection and processing of en-route meteorological data relies on high-precision satellite positioning results for spatiotemporal correlation, the positioning deviations caused by this multipath interference will directly lead to inconsistencies between meteorological parameters and actual track positions, resulting in data spatial mapping misalignment, which in turn affects the input accuracy of subsequent meteorological forecast models and the rationality of path planning, seriously weakening the reliability and responsiveness of the aircraft's flight control system and en-route meteorological warning mechanism.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for collecting and processing meteorological data of low-altitude aircraft routes. By introducing a disturbance feature extraction mechanism based on carrier phase change behavior and combining it with a deep learning model to adaptively predict multipath risks, dynamic solution weight adjustment of multipath contaminated satellites is achieved, thereby enhancing the system's anti-interference ability to environmental changes and the robustness of data processing without completely eliminating the observation source. It not only significantly reduces the risk of spatial mismatch and meteorological data misjudgment caused by multipath, but also ensures the continuous positioning accuracy and track perception reliability of aircraft in high-reflection risk areas such as urban canyons and mountain edges, providing a stable data foundation and technical support for high-precision route meteorological perception, so as to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for collecting and processing meteorological data of a low-altitude aircraft route, comprising the following steps:

[0008] Obtain the carrier phase observation value of the target satellite from the GNSS receiver and record the corresponding timestamp;

[0009] Eliminate the original turnover ambiguity of the acquired carrier phase observation value to ensure the continuity of the carrier phase change;

[0010] First-order difference processing is performed on the carrier phase observations of the same satellite at two consecutive time points to construct an instantaneous carrier phase change sequence that reflects the phase change rate. Based on this instantaneous carrier phase change sequence, feature engineering techniques are used to extract key disturbance features that can characterize potential multipath effects in satellite signals. Within a set analysis time window, the extracted features are quantified and the processed feature indicators are used as feature vectors for subsequent intelligent identification and classification of multipath effect risks.

[0011] The feature vector is input into a deep learning model that has been pre-trained with a large amount of data, and the deep learning model is used to intelligently predict the multipath effect risk of satellite signals;

[0012] When a satellite is identified as having a multipath effect risk, the solution weight of the multipath effect satellite is reduced in real time according to the risk level of the multipath effect, the influence of high-confidence satellites is enhanced, the interference satellites are effectively suppressed, the accuracy of overall positioning and spatial registration of meteorological data is improved, and the data dislocation and misjudgment problems caused by multipath are solved.

[0013] Preferably, the carrier phase observation value of the target satellite is extracted from the GNSS receiver, and the specific steps are as follows:

[0014] First, determine the PRN number of the target satellite and configure the GNSS receiver to record complete observation data (including carrier phase, pseudorange, signal-to-noise ratio, etc.), ensuring that phase observation output for frequencies such as L1 and L2 is enabled.

[0015] Secondly, use the software that comes with the receiver or use standard communication protocols (such as NMEA, UBX, or manufacturer-defined protocols) to export or store the observation data in real time locally;

[0016] Then, parse the format of the exported observation data (e.g., extract the .L1 or .L2 field using the RINEX parsing tool) and locate the phase observation value corresponding to the target PRN number at the corresponding timestamp;

[0017] Finally, the phase value and its corresponding UTC timestamp are recorded as a time series data pair and sorted in sequence according to the sampling frequency to form a continuous carrier phase time series, providing basic data support for subsequent multipath effect identification and analysis.

[0018] Preferably, the original turnaround ambiguity is resolved on the acquired carrier phase observation value, and the specific steps are as follows:

[0019] First, the continuous carrier phase observations of the target satellite are extracted from the GNSS receiver, and the sampling interval and timestamp are checked for integrity to ensure that the data is continuous and without serious omissions.

[0020] Secondly, a sliding window difference analysis is performed on the observation sequence to calculate the carrier phase change between adjacent epochs and preliminarily identify the phase jump point (cycle slip). If the phase change at a certain epoch exceeds the set threshold (such as more than 0.05 cycles), it is marked as a possible cycle anomaly point.

[0021] Then, the ambiguity of the jump point is estimated in integer cycles through dual-frequency phase combination (such as generalized linear combination method) or ambiguity fixing algorithm (such as LAMBDA method), and the integer cycle value of the corresponding epoch is repaired;

[0022] Then, the entire phase time series is filtered and fitted using Kalman filtering or smoothing methods to eliminate small-scale fluctuations and reconstruct a continuous trajectory.

[0023] Finally, the consistency of the corrected phase sequence is verified to ensure that the phase change trend after processing is physically reasonable, continuous and smooth, and has no obvious jumps, thus completing the preliminary ambiguity resolution process.

[0024] Preferably, feature engineering technology is applied to extract key disturbance features that can characterize potential multipath effects in satellite signals, wherein the extracted features include the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence and the energy distribution breadth of the phase change sequence in the spectrum. Within a set analysis time window, the extracted features are quantified to generate a disturbance asymmetry index and a disturbance spectrum dispersion index, respectively. The disturbance asymmetry index and the disturbance spectrum dispersion index are used as feature vectors for subsequent intelligent identification and classification judgment of multipath effect risks.

[0025] Preferably, the disturbance asymmetry index and the disturbance spectrum dispersion index are input as feature vectors into a deep learning model that has been trained in advance with a large amount of data, and the multipath risk coefficient is output by the deep learning model. The multipath effect risk of the satellite signal is intelligently predicted by the multipath risk coefficient.

[0026] Preferably, when a satellite is identified as having a multipath effect risk, the calculation weight of the multipath effect satellite is reduced in real time according to the risk level of the multipath effect. The specific steps are as follows:

[0027] When a satellite is identified as having a multipath effect risk (i.e., the multipath risk coefficient is greater than the multipath risk coefficient reference threshold), a comparison is performed between the multipath risk coefficient output by the deep learning model and the preset multipath risk coefficient reference threshold. In order to achieve a quantitative response to the risk level, a weight attenuation factor is introduced. The calculation expression of the weight attenuation factor is:

[0028] ,in: For the p The multipath risk coefficient of a satellite indicates the intensity of multipath interference to the satellite signal. The larger the value, the higher the risk. This is the reference threshold for the multipath risk factor. The system sets this threshold based on actual scenarios (such as urban environments and open areas) and is used to classify risk levels. is the risk sensitivity coefficient, which controls the response of the multipath risk coefficient to weight attenuation. The larger the value, the more sensitive the system is to risk. For the p The weight attenuation factor of a satellite is used to scale the original observation weight. The smaller the value, the lower the credibility of the satellite observation value.

[0029] After obtaining the weight attenuation factor of each satellite with multipath risk, combined with its original pseudorange observation weight, the adjusted multipath adaptive solution weight is calculated. The specific calculation expression is as follows:

[0030] ,in: For the pThe original observation weight of a satellite when it is not interfered with is usually determined by the satellite elevation angle, signal-to-noise ratio, geometric distribution, etc. It is the actual weight of participating in the solution after adjustment, and is used in positioning algorithms such as Kalman filtering and least squares solution.

[0031] Preferably, within a set analysis time window, the specific steps of quantifying the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence to generate a disturbance asymmetry index are as follows:

[0032] In the analysis time window, the instantaneous carrier phase variation sequence is divided into two sets: positive disturbance set With negative perturbation set , respectively represent the absolute perturbation value sets in the positive and negative directions, Indicates the i The instantaneous carrier phase change at each time point is calculated; secondly, the exponential and transformation cumulative amount of the disturbance amplitude in the two sets are calculated respectively, which are defined as follows: , is the disturbance enhancement coefficient, which is used to emphasize the dominant role of large-amplitude disturbances in asymmetry; is the cumulative value of the positive energy weight in the positive perturbation direction under exponential weighting; is the cumulative value of negative energy weight in the negative disturbance direction under exponential weighting; finally, the disturbance asymmetry index is calculated based on the cumulative value of positive energy weight and the cumulative value of negative energy weight. The calculation expression is: ,in: is a slightly positive number (such as ), used to avoid numerical instability caused by the denominator being zero; It is the disturbance asymmetry index of the final output. The farther the value deviates from 0, the stronger the asymmetry.

[0033] Preferably, within a set analysis time window, the specific steps of quantifying the energy distribution breadth of the phase change sequence in the spectrum to generate a disturbance spectrum dispersion index are as follows:

[0034] Perform Fourier transform on the instantaneous carrier phase variation sequence of the target satellite to obtain its frequency domain amplitude spectrum ,in Indicates the k The amplitude corresponding to each frequency point is N Represents the total number of frequency points; then, construct the normalized spectral energy distribution function , which represents the proportion of each frequency point energy in the overall spectrum; then the disturbance spectrum dispersion index is calculated based on the normalized spectrum energy distribution function, and the calculation expression is:

[0035] ,in: For the kNormalized energy density at each frequency point; j Is the spectrum index variable, used to traverse all frequency domain amplitude spectra; is the disturbance spectrum dispersion index.

[0036] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0037] The present invention introduces a disturbance feature extraction mechanism based on carrier phase change behavior, and combines it with a deep learning model to adaptively predict multipath risks, thereby realizing dynamic solution weight adjustment for multipath contaminated satellites. This enhances the system's anti-interference ability to environmental changes and the robustness of data processing without completely eliminating the observation source. It not only significantly reduces the risk of spatial mismatch and meteorological data misjudgment caused by multipath, but also ensures the continuous positioning accuracy and track perception reliability of aircraft in high-reflection risk areas such as urban canyons and mountain edges, providing a stable data foundation and technical support for high-precision en-route meteorological perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 The present invention is a method flow chart of the low-altitude aircraft route meteorological data collection and processing method. DETAILED DESCRIPTION

[0040] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0041] The present invention provides Figure 1 The method for collecting and processing meteorological data of low-altitude aircraft routes shown includes the following steps:

[0042] Obtain the carrier phase observation value of the target satellite from the GNSS receiver and record the corresponding timestamp;

[0043] Extracting carrier phase observations of a target satellite from a GNSS receiver is typically accomplished through the receiver's raw observation files (e.g., RINEX format) or device interface protocols. First, determine the target satellite's PRN number and configure the GNSS receiver to record complete observation data (including carrier phase, pseudorange, signal-to-noise ratio, etc.), ensuring that phase observation output for frequencies such as L1 and L2 is enabled. Second, use the receiver's accompanying software or standard communication protocols (e.g., NMEA, UBX, or manufacturer-defined protocols) to export or store the observation data in real time locally. Then, parse the exported observation data (e.g., extract the .L1 or .L2 fields using a RINEX parsing tool) to locate the phase observation corresponding to the target PRN number at the corresponding timestamp. Finally, record this phase value and its corresponding UTC timestamp as a time series data pair and organize them sequentially by sampling frequency to form a continuous carrier phase time series, providing the basic data support for subsequent multipath effect identification and analysis.

[0044] Eliminate the original turnover ambiguity of the acquired carrier phase observation value to ensure the continuity of the carrier phase change;

[0045] The original cycle ambiguity resolution of the acquired carrier phase observations can be performed in the following steps: First, continuous carrier phase observations of the target satellite are extracted from the GNSS receiver, and the sampling interval and timestamp are checked for integrity to ensure that the data is continuous and without serious omissions. Second, a sliding window difference analysis is performed on the observation sequence to calculate the carrier phase change between adjacent epochs and preliminarily identify phase slip points. If the phase change at a certain epoch exceeds a set threshold (e.g., more than 0.05 cycles), it is marked as a possible cycle anomaly. Next, the ambiguity of the slip point is estimated by a dual-frequency phase combination (e.g., generalized linear combination method) or an ambiguity fixing algorithm (e.g., LAMBDA method) to repair the integer cycle value of the corresponding epoch. Then, the entire phase time series is filtered and fitted using Kalman filtering or smoothing methods to eliminate small-scale fluctuations and reconstruct a continuous trajectory. Finally, the corrected phase sequence is verified for consistency to ensure that the processed phase change trend is physically reasonable, continuous and smooth, and without obvious slips, thus completing the preliminary ambiguity resolution process.

[0046] The purpose of this step is to convert the original phase observation data from a state containing whole-cycle uncertainty into a continuous, physically interpretable data sequence that can be used for precise solution and interference detection.

[0047] First-order difference processing is performed on the carrier phase observations of the same satellite at two consecutive time points to construct an instantaneous carrier phase change sequence that reflects the phase change rate. Based on this instantaneous carrier phase change sequence, feature engineering techniques are used to extract key disturbance features that can characterize potential multipath effects in satellite signals. Within a set analysis time window, the extracted features are quantified and the processed feature indicators are used as feature vectors for subsequent intelligent identification and classification of multipath effect risks.

[0048] Feature engineering technology is applied to extract key disturbance features that can characterize the potential multipath effects in satellite signals. The extracted features include the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence and the energy distribution breadth of the phase change sequence in the spectrum. Within the set analysis time window, the extracted features are quantified to generate disturbance asymmetry index and disturbance spectrum dispersion index, respectively. The disturbance asymmetry index and disturbance spectrum dispersion index are used as feature vectors for subsequent intelligent identification and classification of multipath effect risks.

[0049] A higher skewness ratio of the absolute values of positive and negative perturbations in the instantaneous carrier phase variation sequence generally indicates a higher risk of multipath interference for the satellite signal. This is because, under ideal conditions, carrier phase variations should exhibit relatively smooth and approximately symmetrical fluctuations with the geometric distance between the satellite and the receiver, with positive and negative variations roughly equal in magnitude. However, when multipath exists, superimposed reflected signals can introduce phase interference in specific directions. For example, during specific periods of time, they can enhance or offset the phase variation of the direct signal, causing the perturbation to skew in one direction. This disrupts the positive-negative symmetry of the phase variation and results in a skewed distribution of the phase variation. Calculating the skewness ratio of the absolute values of positive and negative perturbations (e.g., the ratio of the cumulative amplitude of the positive extreme to the cumulative amplitude of the negative extreme) can effectively reveal whether there is phase imbalance in the signal caused by directional reflections. Therefore, this parameter can serve as a sensitive indicator for identifying multipath interference.

[0050] The specific steps for quantifying the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence within the set analysis time window to generate the disturbance asymmetry index are as follows:

[0051] In the analysis time window, the instantaneous carrier phase variation sequence is divided into two sets: positive disturbance set With negative perturbation set , respectively represent the absolute perturbation value sets in the positive and negative directions, Indicates the i The instantaneous carrier phase change at each time point is calculated; secondly, the exponential and transformation cumulative amount of the disturbance amplitude in the two sets are calculated respectively, which are defined as follows: , is the disturbance enhancement coefficient, which is used to emphasize the dominant role of large-amplitude disturbances in asymmetry; is the cumulative value of the positive energy weight in the positive perturbation direction under exponential weighting; is the cumulative value of negative energy weight in the negative disturbance direction under exponential weighting; finally, the disturbance asymmetry index is calculated based on the cumulative value of positive energy weight and the cumulative value of negative energy weight. The calculation expression is: ,in: is a slightly positive number (such as ), used to avoid numerical instability caused by the denominator being zero; It is the disturbance asymmetry index of the final output. The farther the value deviates from 0, the stronger the asymmetry.

[0052] The disturbance asymmetry index, generated by quantifying the skewness ratio of the absolute values of positive and negative perturbations in the phase variation sequence within a set analysis time window, indicates that larger values indicate a higher risk of multipath for the satellite, while smaller values indicate a lower risk. This is because, in an ideal state without multipath interference, small perturbations in the carrier phase primarily originate from receiver noise or stationary motion between the satellite and the receiver. These variations are evenly distributed in the positive and negative directions, resulting in roughly symmetrical amplitudes and frequencies of positive and negative perturbations. However, when multipath exists, the superimposed reflected signals often exhibit directionality and path dependence, resulting in a persistent shift in one direction (e.g., positive enhancement or negative cancellation), leading to a significant asymmetry in the distribution of perturbation energy in the positive and negative directions. The disturbance asymmetry index, by exponentially enhancing and quantifying this asymmetric structure, reveals the inherent relationship between signal skewness and multipath reflections, making it an important criterion for determining the severity of multipath risk.

[0053] If the energy distribution width of the instantaneous carrier phase change sequence of the same satellite in the spectrum is higher, it usually indicates a higher risk of multipath effects. This is because, ideally, the carrier phase changes of GNSS signals should exhibit relatively stable low-frequency characteristics, reflecting the smooth change of the geometric distance between the satellite and the receiver. However, when multipath effects exist, the direct signal is superimposed on one or more reflected signals. The reflected signals have delays and phase differences, which will introduce high-frequency disturbances and rapid oscillations in the received signal. Such disturbances cause more high-frequency components to appear in the phase change sequence, resulting in the distribution of spectral energy in a wider frequency band, that is, the "energy distribution width" of the spectrum increases. Therefore, the phenomenon of spectrum discretization and expansion is usually a typical frequency domain feature of multipath signal aliasing, and can serve as an important criterion for identifying potential multipath interference.

[0054] The specific steps for quantifying the energy distribution width of the phase change sequence in the spectrum to generate the disturbance spectrum dispersion index within the set analysis time window are as follows:

[0055] Perform Fourier transform on the instantaneous carrier phase variation sequence of the target satellite to obtain its frequency domain amplitude spectrum ,in Indicates the k The amplitude corresponding to each frequency point is N Represents the total number of frequency points; then, construct the normalized spectral energy distribution function , which represents the proportion of each frequency point energy in the overall spectrum; then the disturbance spectrum dispersion index is calculated based on the normalized spectrum energy distribution function, and the calculation expression is:

[0056] ,in: For the k Normalized energy density at each frequency point; j Is the spectrum index variable, used to traverse all frequency domain amplitude spectra; is the disturbance spectrum dispersion index;

[0057] This formula is essentially the energy distribution entropy, which is used to measure the degree of energy diffusion in the spectrum. When the energy is concentrated in a few frequency points (i.e., low frequency is dominant), the disturbance spectrum dispersion index Small, indicating that the signal is stable; when the spectrum energy is widely distributed at multiple frequency points, the disturbance spectrum dispersion index An increase indicates that the spectrum is discrete, the disturbance is complex, and the multipath risk is high.

[0058] The perturbation spectrum dispersion index (DSI) shows that within a set analysis time window, a perturbation asymmetry index (DSI) is generated by quantifying the skewness ratio of the absolute values of positive and negative perturbations in the phase variation sequence. A higher value indicates a higher risk of multipath interference for the satellite, while a lower value indicates a lower risk of multipath interference. This is because, in an ideal, interference-free state, the instantaneous variations in GNSS carrier phase should fluctuate around a stable trajectory, with the amplitudes of positive and negative perturbations roughly symmetrical. However, when multipath occurs, the superposition of reflected and direct signals can create directional shifts or asymmetric interference patterns. For example, some reflected paths consistently lag or lead the phase of the main signal, making the perturbations more likely to be concentrated in a certain direction (positive or negative), thus disrupting the symmetry of the perturbation distribution. By constructing a skewness ratio based on the ratio of the sum of the absolute values of positive and negative perturbations, this index can sensitively capture the phase shift tendency and directional imbalance of perturbations caused by multipath. Therefore, the further the DSI value deviates from 1, the more significant the asymmetric interference effect on the signal, and the higher the multipath risk.

[0059] The feature vector is input into a deep learning model that has been pre-trained with a large amount of data, and the deep learning model is used to intelligently predict the multipath effect risk of satellite signals;

[0060] The disturbance asymmetry index and the disturbance spectrum dispersion index are input as feature vectors into a deep learning model that has been pre-trained with a large amount of data. The multipath risk coefficient is output by the deep learning model, and the multipath effect risk of the satellite signal is intelligently predicted by the multipath risk coefficient.

[0061] A trained deep learning model is an intelligent recognition model that can autonomously identify the characteristics and patterns of multipath disturbances. This model is built using a large amount of labeled GNSS satellite signal carrier phase observation data and a deep learning algorithm (such as a convolutional neural network (CNN), a long short-term memory (LSTM) network, or a temporal transformer model). The core goal of the training process is to enable the model to learn the mapping relationship between disturbances and risks from the input feature vectors, thereby achieving high generalization capabilities for new inputs and quickly outputting the corresponding multipath risk coefficient. During this process, the model gradually fits the complex nonlinear relationship between input features (such as disturbance asymmetry and disturbance spectrum dispersion) and corresponding labels (such as the presence of multipath and the degree of multipath severity) through repeated iterative training, error backpropagation, and parameter adjustment.

[0062] Using the disturbance asymmetry index and disturbance spectrum dispersion index as key input features, these two parameters reveal potential abnormal disturbance behavior in satellite signals from the perspectives of the time and frequency domains, respectively. The disturbance asymmetry index primarily measures the asymmetry between positive and negative fluctuations in the instantaneous carrier phase variation. Normally, the phase variation of a direct signal should appear statistically as a nearly symmetrical noise disturbance. However, once affected by multipath reflections, the uneven direction and intensity of the reflected signals can cause the phase variation sequence to exhibit a significant "skewness" within a certain period of time, such as long-lasting positive or negative disturbances or a significant imbalance in the ratio of upper and lower amplitudes. Extracting the disturbance asymmetry index can sensitively capture this "disturbance skewness," providing the model with important features that reflect directional multipath interference. The disturbance spectrum dispersion index, on the other hand, analyzes the signal energy distribution from a frequency domain perspective. It uses Fourier transform or wavelet transform to analyze the phase disturbance spectrum within the current time window and calculate the distribution breadth of the spectral energy. Because multipath signals are formed by the superposition of multiple delayed paths, they often introduce additional high-frequency components or frequency drift, causing the spectral energy to disperse, spread, or deviate. The perturbation spectrum dispersion metric, by measuring the degree of spectral divergence, can reveal abnormal frequency domain patterns caused by multipath perturbations, providing a deeper description of the perturbation structure for the model.

[0063] The trained deep learning model is built on the basis of large-scale sampling, classification, and mapping of these disturbance features. During the model training phase, developers typically collect satellite signal observation data covering a variety of scenarios (such as open areas, urban canyons, densely built-up areas, and highly reflective surfaces), label this data (e.g., "no multipath," "mild multipath," and "severe multipath"). This labeled data is then used as the training set to construct a supervised learning framework. The model's goal is to minimize the discrepancy between the predicted output (i.e., the multipath risk factor) and the true label, for example, by optimizing using loss functions such as mean squared error (MSE) and cross-entropy. During training, the model continuously improves its ability to identify multipath disturbances by repeatedly learning input features such as the disturbance asymmetry index and spectral dispersion index, and by updating structural weights. In particular, using model structures with sequence modeling capabilities, such as LSTM or Transformer, can further capture the temporal evolution of disturbance features, improving the accuracy of identifying "hidden" or "intermittent" multipath phenomena.

[0064] Ultimately, after training, the model can be deployed in actual flight missions. When new GNSS carrier phase observations are received and the perturbation asymmetry and perturbation spectrum dispersion features are extracted, they are immediately input into the model, which then infers and outputs a corresponding "multipath risk coefficient." This multipath risk coefficient is a normalized value (e.g., a continuous value between 0 and 1) that represents the likelihood and intensity of multipath interference to the current satellite signal. A larger value indicates a higher likelihood and intensity of multipath interference to the current satellite signal, and a higher risk of multipath effects. Conversely, a lower value indicates a lower likelihood and intensity of multipath interference to the current satellite signal, and a lower risk of multipath effects.

[0065] In short, the trained deep learning model not only encodes the complex nonlinear mapping between disturbance characteristics and multipath risk but also serves as a key support module for intelligent GNSS anti-interference and high-precision calibration of en-route meteorological data. By incorporating highly expressive and physically interpretable features such as the "disturbance asymmetry index" and the "disturbance spectrum dispersion index" into the model, combined with deep learning's ability to abstract underlying patterns from massive samples, a significant advancement has been achieved from traditional rule-based filtering to data-driven, adaptive anti-interference.

[0066] The multipath risk coefficient generated by the deep learning model when intelligently predicting the multipath effect risk of satellite signals is compared with the pre-set multipath risk coefficient reference threshold to identify satellites with multipath effect risks. The specific steps are as follows:

[0067] If the multipath risk coefficient is greater than the multipath risk coefficient reference threshold, a risk signal is generated, and it is determined that the satellite has a multipath effect risk; if the multipath risk coefficient is less than or equal to the multipath risk coefficient reference threshold, a normal signal is generated, and it is determined that the satellite does not have a multipath effect risk.

[0068] When a satellite is identified as having a multipath risk, the calculation weight of the multipath-affected satellite is reduced in real time according to the risk level of the multipath effect, thereby enhancing the influence of high-confidence satellites, effectively suppressing interfering satellites, improving the accuracy of overall positioning and spatial registration of meteorological data, and solving the data dislocation and misjudgment problems caused by multipath.

[0069] When a satellite is identified as having a multipath risk, the weight of the satellite is reduced in real time based on the risk level of the multipath effect. The specific steps are as follows:

[0070] When a satellite is identified as having a multipath effect risk (i.e., the multipath risk coefficient is greater than the multipath risk coefficient reference threshold), a comparison is performed between the multipath risk coefficient output by the deep learning model and the preset multipath risk coefficient reference threshold. In order to achieve a quantitative response to the risk level, a weight attenuation factor is introduced. The calculation expression of the weight attenuation factor is:

[0071] ,in: For the p The multipath risk coefficient of a satellite indicates the intensity of multipath interference to the satellite signal. The larger the value, the higher the risk. This is the reference threshold for the multipath risk factor. The system sets this threshold based on actual scenarios (such as urban environments and open areas) and is used to classify risk levels. is the risk sensitivity coefficient, which controls the response of the multipath risk coefficient to weight attenuation. The larger the value, the more sensitive the system is to risk. For the p The weight attenuation factor of a satellite is used to scale the original observation weight. The smaller the value, the lower the credibility of the satellite observation value.

[0072] This step constructs an adaptive, continuously differentiable attenuation curve through an exponential function to ensure that the weight decreases slowly when the risk factor slightly exceeds the standard, while high risk values lead to rapid attenuation, avoiding observation mutations caused by the traditional "hard elimination" mechanism and enhancing system stability.

[0073] After obtaining the weight attenuation factor of each satellite with multipath risk, combined with its original pseudorange observation weight, the adjusted multipath adaptive solution weight is calculated. The specific calculation expression is as follows:

[0074] ,in: For the pThe original observation weight of a satellite when it is not interfered with is usually determined by the satellite elevation angle, signal-to-noise ratio, geometric distribution, etc. The actual weight of the solution after adjustment is used in positioning algorithms such as Kalman filtering and least squares solution;

[0075] This adjustment is performed dynamically in real time and can be updated in each epoch. By integrating risk information with traditional solution factors, a "retention suppression" strategy is implemented: the disturbed satellites are not eliminated, but their influence in the solution system is fine-tuned according to the degree of risk, thereby taking into account both anti-interference and positioning stability. It is suitable for complex scenarios with a limited number of satellites or multiple satellites being disturbed.

[0076] By introducing a dynamic weight adjustment mechanism based on the degree of multipath risk, satellite signals with multipath effect risks are refined and suppressed to prevent their unreliable observation data from misleading the overall positioning results. By reducing the participation weight of high-risk satellites in real time and enhancing the dominant position of undisturbed high-confidence satellites in the solution, it can effectively alleviate problems such as coordinate offset, track distortion, or spatial data mismatch caused by multipath interference, thereby significantly improving the robustness and spatial reference accuracy of GNSS solution results. This strategy not only helps to ensure the stability of aircraft positioning accuracy, but also improves the spatial consistency between meteorological data collected in-situ and the actual flight trajectory, providing more reliable data support for route meteorological modeling, environmental perception, and early warning mechanisms.

[0077] Through the above scheme, it is possible to achieve fine identification and intelligent suppression of GNSS multipath interference for low-altitude aircraft in complex terrain environments, effectively improving the spatial registration accuracy and reliability of route meteorological data. Specifically, the scheme introduces a disturbance feature extraction mechanism based on carrier phase change behavior, and combines it with a deep learning model to adaptively predict multipath risks, thereby achieving dynamic solution weight adjustment for multipath-contaminated satellites. This enhances the system's ability to resist environmental changes and the robustness of data processing without completely eliminating the observation source. This method not only significantly reduces the risk of spatial mismatch and meteorological data misjudgment caused by multipath, but also ensures the continuous positioning accuracy and track perception reliability of aircraft in high-reflection risk areas such as urban canyons and mountain edges, providing a stable data foundation and technical support for high-precision route meteorological perception.

[0078] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0079] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0080] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for collecting and processing meteorological data for low-altitude aircraft routes, characterized in that: The following steps are involved: Obtain the carrier phase observation value of the target satellite from the GNSS receiver, record the corresponding timestamp, and perform raw turnaround ambiguity resolution on the obtained carrier phase observation value; Perform first-order difference processing on the carrier phase observations of the same satellite at two consecutive time points to construct an instantaneous carrier phase change sequence that reflects the phase change rate; Feature engineering techniques are used to extract key disturbance features that can characterize the potential multipath effects of satellite signals from the instantaneous carrier phase variation sequence. Within a set analysis time window, the extracted features are quantified and used as feature vectors for subsequent intelligent identification and classification of multipath effect risks. The feature vector is input into a deep learning model that has been pre-trained with a large amount of data, and the deep learning model is used to intelligently predict the multipath effect risk of satellite signals; When a satellite is identified as having a multipath effect risk, the calculation weight of the multipath effect satellite is reduced in real time according to the risk level of the multipath effect, and the influence of high-confidence satellites is enhanced; Feature engineering technology is applied to extract key disturbance features that can characterize the potential multipath effects in satellite signals. The extracted features include the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence and the energy distribution breadth of the phase change sequence in the spectrum. Within the set analysis time window, the extracted features are quantified to generate disturbance asymmetry index and disturbance spectrum dispersion index, respectively. The disturbance asymmetry index and disturbance spectrum dispersion index are used as feature vectors for subsequent intelligent identification and classification of multipath effect risks.

2. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 1, characterized in that: Extract the carrier phase observation value of the target satellite from the GNSS receiver. The specific steps are as follows: Determine the PRN number of the target satellite and configure the GNSS receiver to record complete observation data and enable phase observation output; The observation data is exported in real time through the standard communication protocol, the format of the exported observation data is parsed, and the phase observation value corresponding to the target PRN number at the corresponding timestamp is located; The phase observation values and their corresponding UTC timestamps are recorded as time series data pairs and sorted in sequence according to the sampling frequency to form a continuous carrier phase time series.

3. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 1, characterized in that: The original turnaround ambiguity is resolved for the acquired carrier phase observations. The specific steps are as follows: Extract continuous carrier phase observations of the target satellite from the GNSS receiver and perform integrity checks on the sampling interval and timestamp; Perform sliding window difference analysis on the observation sequence to calculate the carrier phase change between adjacent epochs and preliminarily identify the phase jump points. If the phase change at a certain epoch exceeds the set threshold, it is marked as a potential turnover anomaly point. The ambiguity fixing algorithm is used to estimate the ambiguity of the jump point in integer cycles and repair the integer cycle value of the corresponding epoch; The Kalman filter method is used to filter and fit the entire phase time series to eliminate small-scale fluctuations and reconstruct a continuous trajectory; The consistency of the corrected phase sequence is verified to ensure that there is no obvious jump in the phase change trend after processing, thus completing the initial ambiguity resolution process.

4. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 1, characterized in that: The disturbance asymmetry index and the disturbance spectrum dispersion index are input as feature vectors into a deep learning model that has been pre-trained with a large amount of data. The multipath risk coefficient is output by the deep learning model, and the multipath effect risk of the satellite signal is intelligently predicted by the multipath risk coefficient.

5. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 4, characterized in that: When a satellite is identified as having a multipath risk, the weight of the satellite is reduced in real time based on the risk level of the multipath effect. The specific steps are as follows: When a satellite is identified as having a multipath effect risk, the multipath risk coefficient output by the satellite is compared with the preset multipath risk coefficient reference threshold, and a weight attenuation factor is introduced to quantitatively respond to the risk level. After obtaining the weight attenuation factor of each satellite with multipath risk, the adjusted multipath adaptive solution weight is calculated in combination with its original pseudorange observation weight.

6. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 1, characterized in that: The specific steps for quantifying the skewness ratio of the absolute values of positive and negative disturbances in the phase change sequence within the set analysis time window to generate the disturbance asymmetry index are as follows: Within the analysis time window, the instantaneous carrier phase change sequence is divided into two sets: a positive disturbance set and a negative disturbance set, which represent the sets of absolute disturbance values in the positive and negative directions, respectively. Secondly, the exponential and transformation cumulative values of the disturbance amplitudes in the two sets are calculated respectively. Finally, the disturbance asymmetry index is calculated based on the positive energy weight cumulative value and the negative energy weight cumulative value.

7. The method for collecting and processing low-altitude aircraft route meteorological data according to claim 1, characterized in that: The specific steps for quantifying the energy distribution width of the phase change sequence in the spectrum to generate the disturbance spectrum dispersion index within the set analysis time window are as follows: Perform Fourier transform on the instantaneous carrier phase variation sequence of the target satellite to obtain its frequency domain amplitude spectrum; Construct a normalized spectrum energy distribution function to represent the proportion of each frequency point energy in the overall spectrum; The disturbance spectrum dispersion index is calculated based on the normalized spectral energy distribution function.

Images