Non-modular error correction method and system for satellite high-precision navigation positioning and storage medium

Abstract

The invention discloses a non-modular error correction method and system for satellite high-precision navigation positioning and a storage medium. The method comprises the following steps: extracting an original observation value of a satellite receiver, and obtaining a residual sequence by using a non-difference and non-combination precision single-point positioning method; calculating a non-modular error feature vector of the residual error sequence according to the time domain and frequency domain features, and constructing an environment recognition model according to the non-modular error feature vector; calculating a correlation coefficient between each error factor and the residual sequence according to the pseudo-range and the carrier phase observation value; establishing a non-modeling error correction model according to the correlation coefficient of each error factor and the residual error sequence, and performing fine adjustment on parameters of the non-modeling error correction model according to a deep transfer learning algorithm; and outputting a corrected result according to the adjusted non-modular error correction model. According to the method, the problem that the existing high-precision navigation positioning non-modular error processing method is insufficient in universality and precision in a complex and changeable environment is solved.

Description

Technical Field

[0001] This invention belongs to the field of satellite communication technology, and in particular relates to a non-modular error correction method, system and storage medium for high-precision satellite navigation and positioning. Background Technology

[0002] In Global Navigation Satellite Systems (GNSS), positioning accuracy is highly dependent on the accuracy of observations such as pseudorange and carrier phase. However, these observations are often affected by various factors, including ionospheric delay, satellite clock errors, orbital errors, and non-modeling errors. Among these, non-modeling errors are particularly prominent in high-precision positioning. These errors are difficult to completely eliminate using conventional differential techniques or traditional models, thus becoming a key factor limiting the accuracy and stability of Precise Point Positioning (PPP). Especially in complex environments such as urban canyons, forest areas, and densely populated areas with tall buildings, satellite signals are easily reflected and scattered, resulting in a large amount of non-modeling errors superimposed on pseudorange and phase observations. This can lead to positioning deviations of tens of centimeters or even higher, severely restricting the application effectiveness of BeiDou in scenarios such as surveying, transportation, autonomous driving, and geological disaster monitoring.

[0003] Existing non-model-based error processing methods mainly rely on filtering, frequency domain analysis, or empirical modeling, such as finite impulse response (FIR) filtering, adaptive filtering, and wavelet transform. These methods can mitigate the impact of errors to some extent. However, these methods often assume that error characteristics are relatively stable under different environments and lack the ability to adapt to the non-stationarity of errors across different scenarios. Therefore, when dealing with non-model-based errors that are highly correlated with the environment and exhibit complex and variable behavior, their universality and accuracy are insufficient.

[0004] To address the limitations of existing non-modular error processing methods for high-precision navigation and positioning in complex and variable environments, this invention proposes a non-modular error correction method, system, and storage medium for satellite high-precision navigation and positioning. Summary of the Invention

[0005] This invention proposes a non-modular error correction method, system, and storage medium for high-precision satellite navigation and positioning, in order to at least solve the problems of insufficient universality and accuracy of existing high-precision navigation and positioning non-modular error processing methods in complex and variable environments.

[0006] According to an embodiment of the present invention, a non-modular error correction method for high-precision satellite navigation and positioning is provided, comprising:

[0007] The raw observations from the satellite receiver are extracted and the residual sequence is obtained using the non-differential, non-combined, precise single-point positioning method.

[0008] Calculate the non-modularized error feature vector of the residual sequence based on the time domain and frequency domain characteristics, and use it to construct an environment recognition model;

[0009] The correlation coefficients between each error factor and the residual sequence are calculated based on the pseudorange and carrier phase observations; the error factors include code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error, and multipath error;

[0010] A non-modeled error correction model is established based on the correlation coefficients between each error factor and the residual sequence, and the parameters of the non-modeled error correction model are fine-tuned using a deep transfer learning algorithm.

[0011] The corrected result is output based on the adjusted non-modeled error correction model.

[0012] In a preferred embodiment, the step of extracting the raw observations from the satellite receiver and obtaining the residual sequence using a non-differential, non-combined precise single-point positioning method includes the following steps:

[0013] Acquire raw observation data collected by the receiver under different typical environments; the raw observation data includes pseudorange observations and carrier phase observations;

[0014] The optimal estimated values ​​of the parameters are obtained by solving the original observation values ​​using the non-differential non-combination precise single-point positioning method;

[0015] The residual sequence is calculated based on the optimal estimate and the original pseudorange and carrier phase observations.

[0016] In a preferred embodiment, the step of calculating the non-modularized error feature vector of the residual sequence based on time-domain and frequency-domain characteristics includes the following steps:

[0017] Temporal features of the residual sequence are extracted to describe the statistical properties and temporal correlation of the unmodulated error;

[0018] Frequency domain features of the residual sequence are extracted to describe the distribution characteristics of the unmodulation error in the frequency domain;

[0019] The time-domain and frequency-domain features of the residual sequence are concatenated to form the non-modularized error feature vector of the residual sequence.

[0020] In a preferred embodiment, constructing the environment recognition model includes the following steps:

[0021] The non-modeled error feature vectors under different environments are used as input parameters for XGBoost training;

[0022] By training an environmental recognition model, the system can automatically identify the current environment.

[0023] In a preferred embodiment, the step of calculating the correlation coefficient between each error factor and the residual sequence based on the pseudorange and carrier phase observations includes the following steps:

[0024] The correlation coefficients between each error factor and the pseudorange residuals are calculated based on the mean and residuals of the pseudorange observations.

[0025] The correlation coefficients between each error factor and the carrier phase residual are calculated based on the mean and residual of the carrier phase observations.

[0026] The correlation coefficients between each error factor and the residual sequence are obtained based on the correlation coefficients between each error factor and the pseudorange residual, and the correlation coefficients between each error factor and the carrier phase residual.

[0027] In a preferred embodiment, establishing a non-modeled error correction model based on the correlation coefficients between each error factor and the residual sequence includes:

[0028] The pseudorange non-modulation error correction amount is calculated based on the mean of each error factor and the correlation coefficient between each error factor and the pseudorange residual.

[0029] The carrier phase non-modular error correction amount is calculated based on the mean of each error factor and the correlation coefficient between each error factor and the carrier phase residual.

[0030] Based on the pseudorange non-modular error correction amount and the carrier phase non-modular error correction amount, non-modular error correction models are established under different environments.

[0031] In a preferred embodiment, the step of fine-tuning the parameters of the non-model-based error correction model according to the deep transfer learning algorithm includes the following steps:

[0032] The task loss function is constructed based on the pseudorange and carrier phase residuals and the corresponding non-modular error correction.

[0033] Fine-tuning the parameters of the correction model minimizes the task loss function, resulting in new pseudorange and carrier phase non-modular error correction values.

[0034] In a preferred embodiment, the step of outputting the corrected result based on the adjusted non-modeled error correction model includes the following steps:

[0035] The pseudorange observations are corrected based on the new pseudorange non-modular error correction amount to obtain the corrected pseudorange;

[0036] The corrected carrier phase is obtained by correcting the carrier phase observations based on the new carrier phase non-modular error correction amount.

[0037] The output pseudorange and carrier phase after correction are the results after non-modular error correction.

[0038] According to another embodiment of the present invention, a computer-readable storage medium is also provided, which stores a computer program for electronic data interchange, wherein the computer program causes a computer to perform the above-described non-modular error correction method for high-precision satellite navigation and positioning.

[0039] According to another embodiment of the present invention, a non-modular error correction system for high-precision satellite navigation and positioning is also provided, comprising:

[0040] Satellite receiver;

[0041] processor;

[0042] Memory;

[0043] And one or more programs, wherein the one or more programs are stored in memory and configured to be executed by the processor, the programs causing the computer to perform the above-described non-modular error correction method for high-precision satellite navigation and positioning.

[0044] The advantages of the non-modular error correction method, system, and storage medium for high-precision satellite navigation and positioning of the present invention are as follows:

[0045] (1) In the data processing stage, non-differential non-combined PPP residual sequences based on multi-environment acquisition are introduced, and the XGBoost classification algorithm is used for environment identification and non-modeling error modeling. Compared with the traditional method that only relies on a single environment model, it can fully characterize the features of non-modeling errors under different environments, and significantly improve cross-scene adaptability and the accuracy of non-modeling error modeling.

[0046] (2) A multivariate constraint model is proposed in the non-modular error correction stage. At the same time, five main influencing factors are introduced, namely code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error and multipath error. Adaptive adjustment is achieved through dynamic weighting. Compared with the traditional single-factor modeling method, different error factors can be mutually restrained in the correction process, thereby effectively improving the robustness and stability of error correction.

[0047] (3) The Deep Transfer Learning algorithm is introduced in terms of model update and transfer. When the receiver enters a new scene, the system can quickly call the existing non-model error model and complete the matching correction. Compared with the traditional method of retraining the model, it not only improves the real-time performance, but also enhances the scalability of the model in cross-scene.

[0048] (4) Establish a model that describes the characteristics of non-modular error under different environments, perform non-modular error correction through multivariate constraints, and quickly match and call the non-modular error correction model when updating the model. Compared with the traditional non-modular error correction method, it can effectively reduce the level of non-modular error in pseudorange / carrier phase observations, improve the modeling and correction efficiency of errors, and thus provide more reliable observation data for high-precision positioning. Attached Figure Description

[0049] Figure 1 This is a flowchart of a non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0050] Figure 2 This is a flowchart of step S01 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0051] Figure 3 This is a flowchart of step S02 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0052] Figure 4 This is a flowchart of step S03 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0053] Figure 5 This is a flowchart of step S04 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0054] Figure 6 This is a flowchart of step S05 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0055] Figure 7 This is a flowchart of step S06 of the non-modular error correction method for high-precision satellite navigation and positioning according to an embodiment of the present invention;

[0056] Figure 8 This is a schematic diagram of the structure of a non-modular error correction system for high-precision satellite navigation and positioning according to an embodiment of the present invention. Detailed Implementation

[0057] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention. The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. No specific limitations are made in this application depending on the specific settings of the actual application environment.

[0058] According to an embodiment of the present invention, a non-modular error correction method for high-precision satellite navigation and positioning is provided, the flowchart of which is shown below. Figure 1 As shown, it includes:

[0059] Step S01: Extract the raw observations from the satellite receiver and obtain the residual sequence using the non-differential non-combination precise single-point positioning method;

[0060] Step S02: Calculate the non-modularized error feature vector of the residual sequence based on the time domain and frequency domain characteristics, and use it to construct an environment recognition model;

[0061] Step S03: Calculate the correlation coefficients between each error factor and the residual sequence based on the pseudorange and carrier phase observations; the error factors include code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error, and multipath error;

[0062] Step S04: Establish a non-modeling error correction model based on the correlation coefficients between each error factor and the residual sequence;

[0063] Step S05: Fine-tune the parameters of the non-modeling error correction model according to the deep transfer learning algorithm;

[0064] Step S06: Output the corrected result based on the adjusted non-modeled error correction model.

[0065] In a preferred embodiment, step S01 involves extracting the raw observations from the satellite receiver and obtaining the residual sequence using a non-differential, non-combined precise single-point positioning method, as shown in the flowchart below. Figure 2 As shown, the steps include:

[0066] Step S011: Obtain raw observation data collected by the receiver under different typical environments; the raw observation data includes pseudorange observation values ​​and carrier phase observation values;

[0067] Step S012: Use the non-differential non-combination precise single-point positioning method to solve the original observation values ​​to obtain the optimal estimated values ​​of the parameters;

[0068] Step S013: Calculate the residual sequence based on the optimal estimate and the original pseudorange and carrier phase observations.

[0069] In this embodiment, observation experiments were first conducted in various typical environments, including open fields, urban canyons, forest areas, tunnel entrances, and densely populated high-rise areas. The observation experiments covered multiple time periods and frequencies to ensure data diversity and representativeness. The raw observation data collected by the receiver included pseudorange and carrier phase observations, covering multiple frequencies and constellations of the BeiDou system. The raw observations were processed using the non-differential, non-combined precise point positioning (PPP) method to eliminate major error sources such as ionospheric delay, satellite orbit errors, and receiver clock errors, resulting in a residual sequence. The non-modeling errors remaining in the residual sequence serve as the basis for subsequent modeling and correction. The formulas for the raw pseudorange and carrier phase observations are as follows:

[0070] (1)

[0071] In the formula, Represents pseudorange and carrier phase observations; This represents the geometric distance between the satellite and the phase center of the receiver antenna; Represents the speed of light, which is a known quantity; Indicates receiver clock bias; This represents the satellite clock bias, which is a known quantity. Indicates tropospheric delay; Indicates ionospheric delay; This represents the inter-frequency coefficient of the ionosphere, which is a known quantity; These represent the non-modulation errors in pseudorange and carrier phase observations, respectively; This represents the ambiguity parameter.

[0072] Optimal parameter estimates are obtained using non-difference, non-combination PPP. X = [ ρ ⌢ , Δ t ⌢ r , T ⌢ r , I ⌢ , B ⌢ ] , to the optimal estimate Substituting the original pseudorange and carrier phase observation equation (1), the residual is obtained as follows:

[0073] (2)

[0074] In the formula, This represents the pseudorange and carrier phase observations at the corresponding frequency. This represents the optimal estimate of the geometric distance between the satellite and the phase center of the receiver antenna; Represents the speed of light; This represents the optimal estimate of the receiver clock bias; This represents the optimal estimate of tropospheric delay; The optimal estimate of ionospheric delay; This represents the inter-frequency coefficient of the ionosphere, which is a known quantity. These represent the non-differential, non-combined PPP pseudorange and carrier phase residual, respectively, i.e., the non-modular error.

[0075] In a preferred embodiment, step S02, calculating the non-modularized error feature vector of the residual sequence based on time-domain and frequency-domain features and constructing an environment recognition model, is illustrated in the flowchart below. Figure 3 As shown, the steps include:

[0076] Step S021: Extract the time-domain features of the residual sequence to describe the statistical characteristics and temporal correlation of the non-modeling error;

[0077] Step S022: Extract the frequency domain features of the residual sequence to describe the distribution characteristics of the non-modulation error in the frequency domain;

[0078] Step S023: Concatenate the time-domain features and frequency-domain features of the residual sequence to form the non-modularized error feature vector of the residual sequence;

[0079] Step S024: Use the non-modeled error feature vectors under different environments as XGBoost input parameters for training;

[0080] Step S025: Establish an environment recognition model through training to achieve automatic identification of the current environment.

[0081] In this embodiment, the non-modeling error obtained in step S01 is subjected to time-domain feature processing (including mean and variance), as shown in the following formula:

[0082] (3)

[0083] In the formula, Indicates different observation environments; These represent the average values ​​of pseudorange and carrier phase residuals under the corresponding environments; These represent the variances of the pseudorange and carrier phase residuals, respectively, under the corresponding environments. These represent the number of pseudorange and carrier phase residual samples, respectively, under the corresponding environment. The pseudorange and carrier phase residuals respectively correspond to the following environments. One sample; then, frequency domain feature processing (including power spectrum) is performed, as shown in the following formula:

[0084] (4)

[0085] In the formula, Indicates different observation environments; They represent the first under the corresponding environment. DFT of pseudorange and carrier phase residual samples; They represent the corresponding environments. Power spectrum of pseudorange and carrier phase residual samples; These represent the number of pseudorange and carrier phase residual samples, respectively, under the corresponding environment. The pseudorange and carrier phase residuals respectively correspond to the following environments. One sample; The sampling interval for pseudorange and carrier phase observations is expressed in seconds.

[0086] The above features are concatenated to form a feature vector. :

[0087] Y e = [ μ P , e , μ L , e , σ P , e , σ L , e , S P , e , S L , e ] (5)

[0088] In the formula, Indicates different observation environments; These represent the average values ​​of the pseudorange and carrier phase residuals obtained by formula (3) under the corresponding environment; Let represent the variances of the pseudorange and carrier phase residuals obtained from formula (3) under the corresponding environment, respectively; The power spectrum sequences of pseudorange and carrier phase residuals obtained by formula (4) under the corresponding environment are respectively represented.

[0089] Finally, the feature vectors under different environments were analyzed. The parameters are used as input parameters for XGBoost to train an environment recognition model. XGBoost, as an ensemble learning method based on gradient boosting trees, can effectively handle high-dimensional, non-linear, and complex distributed feature data. During model training, the parameters booster in XGBoost are set to gbtree, max_depth to 8, alpha to 0.3, and lambda to 0.2.

[0090] An environment recognition model is established by training on residual samples of labeled environments. When the receiver enters a new environment, the system can quickly determine its environment category based on the PPP residual characteristics, thus providing accurate prior environmental information for subsequent error modeling and correction.

[0091] In a preferred embodiment, the error factors include code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error, and multipath error. In this embodiment, the code phase error is calculated based on the pseudorange and carrier phase observations collected by the receiver under a certain environment in step S01. Elevation angle error Signal strength error Carrier-to-noise ratio error and multipath error The calculation methods for these five error factors are common knowledge in the field of positioning, and will not be elaborated here.

[0092] In a preferred embodiment, step S03 involves calculating the correlation coefficients between each error factor and the residual sequence based on the pseudorange and carrier phase observations, as shown in the flowchart below. Figure 4 As shown, the steps include:

[0093] Step S031: Calculate the correlation coefficient between each error factor and the pseudorange residual based on the mean and residual of the pseudorange observations;

[0094] Step S032: Calculate the correlation coefficient between each error factor and the carrier phase residual based on the mean and residual of the carrier phase observations;

[0095] Step S033: Obtain the correlation coefficient between each error factor and the residual sequence based on the correlation coefficient between each error factor and the pseudorange residual and the correlation coefficient between each error factor and the carrier phase residual.

[0096] In this embodiment, the correlation coefficients between each error factor and the corresponding non-differential, non-combined PPP residuals obtained by formula (2) are calculated respectively, as follows:

[0097] (6)

[0098] In the formula, These represent the five error factors: code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error, and multipath error, respectively. These represent the correlation coefficients between the corresponding error factors and the pseudorange and carrier phase residuals, respectively. Indicates the number of samples; They represent the first One error factor, pseudorange residual, and carrier phase residual; These represent the mean values ​​of pseudorange and carrier phase residual, respectively.

[0099] In a preferred embodiment, step S04 involves establishing a non-modeled error correction model based on the correlation coefficients between each error factor and the residual sequence, as shown in the flowchart below. Figure 5 As shown, it includes:

[0100] Step S041: Calculate the pseudorange non-modularization error correction amount based on the mean of each error factor and the correlation coefficient between each error factor and the pseudorange residual.

[0101] Step S042: Calculate the carrier phase non-modular error correction amount based on the mean of each error factor and the correlation coefficient between each error factor and the carrier phase residual.

[0102] Step S043: Establish non-mode error correction models under different environments based on pseudorange non-mode error correction amount and carrier phase non-mode error correction amount.

[0103] In this embodiment, an error correction model is established based on the correlation coefficient, as follows:

[0104] (7)

[0105] In the formula, These represent the mean values ​​of code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error, and multipath error in formula (6), respectively. These represent the correlation coefficients between the corresponding error factors and the pseudorange and carrier phase residuals, respectively. These represent the pseudorange and carrier phase non-modulation error correction amounts, respectively.

[0106] In a preferred embodiment, step S05 involves fine-tuning the parameters of the non-model-based error correction model using a deep transfer learning algorithm, as shown in the flowchart below. Figure 6 As shown, the steps include:

[0107] Step S051: Construct the task loss function based on the pseudorange and carrier phase residuals and the corresponding non-modular error correction amount;

[0108] Step S052: Fine-tune the parameters of the correction model to minimize the task loss function, and obtain new pseudorange and carrier phase non-modular error correction amounts.

[0109] In this embodiment, in a practical application scenario, 100 satellite pseudorange and carrier phase observations are first collected. The non-differential, non-combined PPP residuals are calculated according to formulas (2,3,4,5), and features are extracted to obtain the feature vector. :

[0110] Y = [ μ P , μ L , σ P , σ L , S P , S L ] (8)

[0111] In the formula, These represent the average values ​​of the pseudorange and the carrier phase residual, respectively. These represent the variances of the pseudorange and the carrier phase residual, respectively. These represent the power spectrum sequences of pseudorange and phase residual, respectively.

[0112] Then the feature vector Y = [ μ P , μ L , σ P , σ L , S P , S L ] Input the pre-trained XGBoost environment recognition model from step S02 to obtain the current environment category. Based on the environment category output by XGBoost, match it with the pre-established non-model-based error correction model from step S03. Simultaneously, to improve the performance of the pre-established non-model-based error correction model, a Deep Transfer learning algorithm (parameter settings: Optimizer set to Adam, Epochs set to 200, BatchSize set to 32) is introduced to fine-tune the correction model parameters, reducing the task loss... To minimize this, the task loss function is as follows.

[0113] (9)

[0114] In the formula, These represent the original pseudorange and carrier phase non-modulation error correction amounts obtained in step S03, respectively. They represent the first The pseudorange and carrier phase residuals are used to obtain new pseudorange and carrier phase non-modular error correction quantities after fine-tuning the correction model using the Deep Transfer learning algorithm. It can be used for non-modeling error correction in the current application environment.

[0115] In a preferred embodiment, step S06 involves outputting the corrected result based on the adjusted non-modeled error correction model, as shown in the flowchart below. Figure 7 As shown, the steps include:

[0116] Step S061: Correct the pseudorange observations according to the new pseudorange non-modularization error correction amount to obtain the corrected pseudorange;

[0117] Step S062: Correct the carrier phase observation value according to the new carrier phase non-modular error correction amount to obtain the corrected carrier phase;

[0118] Step S063: Output the corrected pseudorange and carrier phase, which is the result after non-modular error correction.

[0119] In this embodiment, the output is the result after non-modularization error correction, and the correction formula is as follows:

[0120] (10)

[0121] In the formula, They represent the first Correction results for pseudorange and carrier phase observations; They represent the first One pseudorange and carrier phase observation; These represent the pseudorange and carrier phase non-modulation error correction amounts obtained in step S04, respectively. This indicates the number of pseudorange and carrier phase observations.

[0122] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, which stores a computer program for electronic data interchange, wherein the computer program causes a computer to perform the non-modular error correction method for high-precision satellite navigation and positioning according to any of the above embodiments.

[0123] According to another embodiment of the present invention, a non-modular error correction system for high-precision satellite navigation and positioning is also provided, the structural schematic diagram of which is shown below. Figure 8 As shown, it includes:

[0124] Satellite receiver;

[0125] processor;

[0126] Memory;

[0127] And one or more programs, wherein the one or more programs are stored in memory and configured to be executed by the processor, the programs causing the computer to perform the non-modular error correction method for high-precision satellite navigation and positioning according to any of the above embodiments.

[0128] The methods described above according to the invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA) for such software processing. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods described herein. Furthermore, when a general-purpose computer accesses the code used to implement the processing shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for performing the processing shown herein.

[0129] Of course, those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Any changes or modifications to the above embodiments that are within the scope of the present invention will fall within the protection scope of the present invention.

Claims

1. A non-modeling error correction method for high-precision navigation and positioning of satellites, characterized in that, The method comprises the steps of: extracting original observation values of a satellite receiver and obtaining residual sequences by using a non-difference non-combination precise point positioning method; calculating non-modeled error feature vectors of the residual sequences according to time domain and frequency domain characteristics and constructing an environment identification model based on the non-modeled error feature vectors; calculating correlation coefficients of each error factor and the residual sequences according to pseudo-range and carrier phase observation values; the error factors include code phase error, elevation angle error, signal strength error, carrier-to-noise ratio error and multipath error; establishing a non-modeled error correction model according to the correlation coefficients of each error factor and the residual sequences; fine-tuning parameters of the non-modeled error correction model according to a deep transfer learning algorithm; outputting corrected results according to the adjusted non-modeled error correction model.

2. The method of claim 1, wherein the non-modeled error correction is performed by a Kalman filter. The method of extracting original observation values of a satellite receiver and obtaining residual sequences by using a non-difference non-combination precise point positioning method comprises the steps of: acquiring original observation data collected by a receiver in different typical environments; the original observation data includes pseudo-range observation values and carrier phase observation values; calculating optimal estimated values of parameters by using a non-difference non-combination precise point positioning method to solve the original observation values; calculating residual sequences according to the optimal estimated values and the original pseudo-range and carrier phase observation values.

3. The method of claim 1, wherein the non-modeled error correction is a function of the satellite high precision navigation positioning system. The method of calculating non-modeled error feature vectors of the residual sequences according to time domain and frequency domain characteristics comprises the steps of: extracting time domain characteristics of the residual sequences to describe statistical characteristics and time correlation of the non-modeled errors; extracting frequency domain characteristics of the residual sequences to describe distribution characteristics of the non-modeled errors in the frequency domain; splicing the time domain characteristics and the frequency domain characteristics of the residual sequences to form the non-modeled error feature vectors of the residual sequences.

4. The method of claim 3, wherein, The method of constructing the environment identification model comprises the steps of: training non-modeled error feature vectors in different environments as XGBoost input parameters; establishing the environment identification model by training to realize automatic discrimination of a current environment.

5. The method of claim 1, wherein, The method of calculating correlation coefficients of each error factor and the residual sequences according to pseudo-range and carrier phase observation values comprises the steps of: calculating correlation coefficients of each error factor and pseudo-range residuals according to the mean value of the pseudo-range observation values and the residuals; calculating correlation coefficients of each error factor and carrier phase residuals according to the mean value of the carrier phase observation values and the residuals; obtaining correlation coefficients of each error factor and the residual sequences according to the correlation coefficients of each error factor and the pseudo-range residuals and the correlation coefficients of each error factor and the carrier phase residuals.

6. The method of claim 1, wherein, The method of establishing a non-modeled error correction model according to the correlation coefficients of each error factor and the residual sequences comprises: calculating pseudo-range non-modeled error correction amounts according to the mean value of each error factor and the correlation coefficients of each error factor and the pseudo-range residuals; calculating carrier phase non-modeled error correction amounts according to the mean value of each error factor and the correlation coefficients of each error factor and the carrier phase residuals; establishing non-modeled error correction models in different environments according to the pseudo-range non-modeled error correction amounts and the carrier phase non-modeled error correction amounts.

7. The method of claim 6, wherein the non-modeled error correction is based on a combination of a Kalman filter and a neural network. The method of fine-tuning parameters of the non-modeled error correction model according to a deep transfer learning algorithm comprises the steps of: ​ constructing a task loss function according to pseudo-range and carrier phase residuals and corresponding non-modeled error correction amounts; The fine tuning of the correction model parameters minimizes the task loss function, and new pseudo-range and carrier phase non-model error correction amounts are obtained.

8. The non-modular error correction method for high-precision satellite navigation and positioning according to claim 7, characterized in that, The step of outputting the corrected results according to the adjusted non-model error correction model comprises: The pseudo-range observation value is corrected according to the new pseudo-range non-model error correction amount to obtain a corrected pseudo-range; The carrier phase observation value is corrected according to the new carrier phase non-model error correction amount to obtain a corrected carrier phase; The output of the corrected pseudo-range and carrier phase is the result of the non-model error correction.

9. A computer readable storage medium storing a computer program for electronic data interchange, wherein, The computer program enables the computer to execute the method of any one of claims 1-8.

10. A non-modelled error correction system for high precision navigation and positioning of satellites, characterized in that Comprise: a satellite receiver; a processor; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs enable the computer to execute the method of any one of claims 1-8.

Images