A GNSS RTK positioning and star selection method and system driven by both model and data
The GNSS RTK positioning method enhances satellite selection using hybrid models and data verification to address computational and error challenges, improving accuracy and efficiency for low-cost devices in urban scenarios.
Patent Information
- Application Number
- CN202211484495.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-24
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2042-11-24
AI Technical Summary
Low-cost GNSS devices face challenges in urban environments due to high computational load and increased gross errors, affecting positioning accuracy and efficiency.
A GNSS RTK positioning method that selects satellites using hybrid models of pseudorange noise, altitude angle, and signal-to-noise ratio, followed by pre- and post-verification residual data to eliminate gross errors and reduce computational load.
Improves RTK positioning accuracy and reduces computational resources, meeting real-time high-precision positioning requirements for low-cost navigation modules.
Smart Images
Figure CN115792980B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of GNSS navigation and positioning, and in particular relates to a GNSS RTK positioning and star selection method and system driven by both a model and data. Background Art
[0002] With the official completion of my country's BeiDou-3 (BDS-3) global system in July 2020, the global satellite navigation system (GNSS) has entered a new era of integrated development of multiple systems such as GPS, BDS, GLONASS, and GALILEO. Currently, there are more than 100 GNSS satellites in orbit providing satellite signals of more than ten different frequencies to users around the world. These emerging satellite navigation systems and rich frequency resources have greatly improved the satellite geometry and brought more observation values, thereby shortening the positioning convergence time, improving the accuracy and reliability of GNSS positioning, and bringing new development opportunities to navigation and positioning.
[0003] Nowadays, the precise positioning services provided by GNSS are gradually moving from limited demand in professional fields to widespread demand among mass users. Low-cost positioning terminals have become the mainstream carriers of navigation and positioning services due to their low cost, low power consumption, and high sensitivity. However, multi-frequency and multi-mode satellite signals and urban-based application scenarios have brought new challenges to low-cost positioning terminals: on the one hand, due to factors such as computing resources and battery capacity, when there are many visible satellites, the dimension of the observation equation of GNSS positioning is large, resulting in a huge consumption of computing resources for matrix operations; on the other hand, due to the limitations of low-cost, low-power GNSS devices and the complex and changeable application environment, a large number of abnormal observations such as gross errors are prone to appear in the original GNSS observations, which not only increases the number of robustness iterations and reduces the efficiency of solution, but also destroys the accuracy and reliability of positioning.
[0004] In order to balance the computational efficiency and positioning performance and meet the demand for real-time high-precision positioning of low-cost navigation modules, the present invention provides a GNSS RTK positioning star selection method driven by both models and data. The pre-selection method is driven by a hybrid model of altitude angle, signal-to-noise ratio and GDOP, the rough selection method is driven by pre-verification residual data, and the fine selection method is driven by post-verification standardized residual data. The method can effectively identify and eliminate gross error observations, thereby improving the RTK positioning accuracy and reducing the computing resource consumption caused by positioning solution. Summary of the invention
[0005] The present invention proposes a GNSS RTK positioning and star selection method driven by both models and data, which can effectively reduce the computational load and improve the computational efficiency while taking into account the positioning performance, and meet the requirements of low-cost, low-power GNSS module real-time solution in terms of computational resources and efficiency. The overall structure of the method of the present invention is as followsFigure 1 As shown, the technical solution is as follows:
[0006] A GNSS RTK positioning satellite selection method driven by both models and data.
[0007] Perform the first satellite elimination: use the hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio to eliminate satellites with poor quality;
[0008] Perform the second satellite elimination: pre-screen each remaining satellite based on its contribution to GDOP;
[0009] The third satellite elimination is performed: for the pre-screened satellites, the pre-verification pseudorange and phase double difference residuals are calculated based on their predicted positions at the current epoch, outliers are eliminated, and the rough screening is completed.
[0010] The fourth satellite elimination is performed: the satellites after rough screening enter the RTK floating point solution, and the satellites are finely screened based on the post-verification standardized residuals and then enter the ambiguity fixation.
[0011] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data, before the first satellite elimination, the hybrid model construction method of altitude angle and signal-to-noise ratio is: collect a large sample of GNSS original observation data covering typical complex scenes in the city, substitute the true position value to extract the true pseudorange noise of each satellite, and thus construct a hybrid model of pseudorange noise, altitude angle and signal-to-noise ratio.
[0012] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data, when the third satellite elimination is performed, the Doppler velocity measurement solution is used to obtain high-precision velocity information using the roughly screened satellites, and the position change equation between carrier epochs is constructed to calculate the predicted position of the current epoch. The predicted position is substituted into the pseudorange double difference and phase double difference observation equations to calculate the pre-test residual, and the Grubbs criterion is used to eliminate outliers in the pre-test residual to complete the rough selection of satellites.
[0013] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data, when the fourth satellite is eliminated, the code phase combination double difference observation value is constructed to estimate the observation noise in real time, and then the variance covariance matrix of the double difference pseudorange / phase is constructed. After the floating point solution is completed, the post-test normalized residual v of each observation value is calculated using the post-test residual and the observation variance, and the ambiguity is fixed after eliminating the satellite with v exceeding the limit.
[0014] In the above-mentioned GNSS RTK positioning star selection method driven by both model and data, the pseudo-range noise of the large sample original observation data is extracted in the following way:
[0015]
[0016] The above formula is the pseudo-range non-difference observation equation, where: is the original pseudorange observation value, ρ represents the satellite-to-earth distance, δt R and δt S Represent the receiver clock error and satellite clock error, δ trop and δ ion represent the tropospheric delay error and the ionospheric delay error, δ mul and ε ρ are the multipath error and pseudorange observation noise, respectively.
[0017] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data, by analyzing the correlation between pseudorange noise and signal-to-noise ratio and altitude angle, a hybrid model of signal-to-noise ratio and altitude angle is constructed to calculate the theoretical noise value. The theoretical noise model δ of satellite i at the fth frequency point i,f The build is as follows:
[0018]
[0019] Among them, q 1 、a 1 , b 1 ,q 2 、a 2 , b 2 is the empirical parameter to be fitted, el i represents the altitude angle of satellite i, SNR i,f represents the signal strength of the fth frequency point of satellite i. Substituting the real noise, signal-to-noise ratio and altitude angle extracted from large sample data into the above formula, six empirical parameters can be fitted.
[0020] When real-time data is obtained, the theoretical noise δ can be calculated by formula (2): i,f , when the theoretical noise satisfies δ i,f ≤δ 0 , which is selected as an available satellite.
[0021] In the above model and data dual-driven GNSS RTK positioning satellite selection method, the contribution of each remaining satellite to GDOP is calculated. The formula is as follows:
[0022]
[0023]
[0024]
[0025] Among them, h j is the observation vector corresponding to the i-th satellite, H nis the observation matrix corresponding to n satellites.
[0026] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data, it is defined that the measuring stations u and r simultaneously observe the GNSS satellites i and j. Then the GNSS pseudorange double difference observation equation and the carrier phase double difference observation equation are as follows:
[0027]
[0028]
[0029] Where P is the pseudorange observation value, λ is the carrier wavelength, is the carrier phase observation value, ρ is the geometric distance between the satellite and the receiver, I and T are the residual errors of ionospheric delay and tropospheric delay respectively. In the short baseline mode, the atmospheric residual error can be ignored, ε is the comprehensive term of pseudorange observation noise and multipath error, and ζ is the comprehensive term of phase observation noise and multipath error.
[0030] In the above-mentioned GNSS RTK positioning satellite selection method driven by both model and data,
[0031] Construct the CMC combined observation value of satellite i and perform inter-station difference and inter-satellite difference on it to obtain the double-difference CMC combined observation value ( Figure 4 Ci in Chinese:
[0032]
[0033] Among them, ▽Δ is the double difference operator, and the double difference ambiguity is a constant term,
[0034] And determine whether a cycle slip occurs. If a cycle slip occurs, let C i is the initial error C 0 , observation noise W i W i-1 After that, construct the observation noise matrix W. Otherwise, construct the CMC triple difference observation C i -C i-1 , and form the observation noise matrix W.
[0035] Kalman filtering is performed based on the constructed observation noise matrix W. After Kalman filtering is completed, the standardized residual v of each phase observation value is calculated s,i , as shown below:
[0036]
[0037] Among them, v i is the posterior residual of phase observation i, σ iThe diagonal elements of the variance matrix of the phase observations constructed by the CMC combined observations, if v s,i <k 0 ,k 0 =1.5, it is selected as the available floating point ambiguity and enters the ambiguity fixing stage. Otherwise, after removing satellite i, Kalman filtering is performed again, the standardized residual is calculated, and it is judged again whether it passes the test.
[0038] A GNSS RTK positioning and star selection system driven by both models and data, including
[0039] The first module is configured to perform the first satellite elimination: using a hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio to eliminate satellites of poor quality;
[0040] The second module is configured to perform the second satellite elimination: pre-screening each remaining satellite according to its contribution to GDOP;
[0041] The third module is configured to perform the third satellite elimination: for the pre-screened satellites, the pre-verification pseudorange and phase double difference residuals are calculated based on their predicted positions at the current epoch, outliers are eliminated, and rough screening is completed.
[0042] The fourth module is configured to perform the fourth satellite elimination: the satellites after coarse screening enter the RTK floating point solution, and the satellites are finely screened based on the post-verification standardized residuals and then enter the ambiguity fixation.
[0043] The GNSS RTK positioning and star selection method driven by both model and data proposed in the present invention has the following advantages:
[0044] 1. By collecting a large sample of GNSS data covering typical complex urban scenes, a hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio is fitted, which improves the degree of consistency between the theoretical value of noise and the true value, thereby facilitating the accurate identification of satellites with large gross errors in urban scenes.
[0045] 2. The position of the previous epoch is transmitted using high-precision velocity measurement information, which improves the accuracy of the predicted position, thereby effectively detecting gross errors through pre-verification residuals.
[0046] 3. By constructing the CMC combined double-difference observation equation, the observation noise can be estimated in real time. The variance-covariance matrix of the observation value constructed in this way can better reflect the actual observation noise, which is beneficial to improving the accuracy of RTK floating-point solution. At the same time, the ambiguity with better quality can be effectively selected through the post-verification standardized residual to enter the ambiguity fixation solution, thereby improving the ambiguity fixation rate.
[0047] 4. By using the GNSS RTK satellite selection method driven by both models and data, higher-quality satellites are selected for GNSS solution. While ensuring positioning performance, the computing load is reduced and the computing rate is improved. This can meet the real-time solution requirements of low-cost navigation modules that are limited by factors such as computing resources and battery capacity. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 The overall flow chart of the GNSS RTK positioning and star selection method driven by both the model and data of the example of the present invention;
[0049] Figure 2 A flow chart of the hybrid model driven pre-selection of stars according to an example of the present invention;
[0050] Figure 3 This is a flow chart of rough star selection driven by pre-test data in an example of the present invention;
[0051] Figure 4 This is a flowchart of the post-test data driven selection of stars according to an example of the present invention. DETAILED DESCRIPTION
[0052] The various modules of the present invention will be described in detail below in conjunction with specific implementation methods.
[0053] 1. Hybrid model driven pre-selection method.
[0054] Carry out vehicle testing of low-cost GNSS navigation modules to cover typical complex environments in cities to the greatest extent possible and obtain large sample GNSS raw observation data. Perform short baseline observations on the GNSS navigation module and extract pseudo-range noise from large sample raw observation data by substituting high-precision reference coordinates into the double-difference model. The extraction method is as follows:
[0055]
[0056] The above formula is the pseudo-range non-difference observation equation, where: is the original pseudorange observation value, ρ represents the satellite-to-earth distance, δt R and δt S Represent the receiver clock error and satellite clock error, δ trop and δ ion represent the tropospheric delay error and the ionospheric delay error, δ mul and ε ρare multipath error and pseudorange observation noise respectively. Perform inter-satellite single difference and inter-station single difference on the non-difference observation equation to eliminate the receiver clock error, satellite clock error, ionospheric delay error and tropospheric delay error, and substitute the high-precision reference coordinates to calculate a more accurate satellite-to-earth distance term. After deducting the above known error terms and satellite-to-earth distance terms, the comprehensive noise term δ including multipath error and pseudorange observation noise can be extracted. By analyzing the correlation between pseudorange noise and signal-to-noise ratio and altitude angle, a hybrid model of signal-to-noise ratio and altitude angle is constructed to calculate the theoretical noise value. The theoretical noise model δ of satellite i at the fth frequency point i,f The build is as follows:
[0057]
[0058] Among them, q 1 、a 1 , b 1 ,q 2 、a 2 , b 2 is the empirical parameter to be fitted, el i represents the altitude angle of satellite i, SNR i,f represents the signal strength of the fth frequency point of satellite i. Substituting the real noise, signal-to-noise ratio and altitude angle extracted from large sample data into the above formula, six empirical parameters can be fitted.
[0059] When real-time data is obtained, the theoretical noise δ can be calculated by formula (2): i,f , when the theoretical noise satisfies δ i,f ≤δ 0 , that is, selected as an available satellite, δ 0 The value of depends on the specific characteristics of the navigation module.
[0060] After preliminary screening by the altitude angle and signal-to-noise ratio hybrid model, the contribution of each remaining satellite to GDOP is calculated. The formula is as follows:
[0061]
[0062]
[0063]
[0064] Among them, h j is the observation vector corresponding to the i-th satellite, H n is the observation matrix corresponding to n satellites. The larger the value, the greater the contribution of the satellite to GDOP. Therefore, the first m satellites after sorting are selected. 2. Pre-test data driven rough star selection method.
[0065] Assuming that stations u and r observe GNSS satellites i and j at the same time, the GNSS pseudorange double difference observation equation and carrier phase double difference observation equation are as follows:
[0066]
[0067]
[0068] Where P is the pseudorange observation value, λ is the carrier wavelength, is the carrier phase observation value, ρ is the geometric distance between the satellite and the receiver, I and T are the residual errors of ionospheric delay and tropospheric delay respectively, in the short baseline mode, the atmospheric residual error can be ignored, ε is the comprehensive term of pseudorange observation noise and multipath error, ζ is the comprehensive term of phase observation noise and multipath error. Therefore, if the error comprehensive term is to be extracted, the approximate position of the receiver needs to be known. The present invention obtains high-precision velocity information through Doppler velocity measurement, passes down the filtered solution of the previous epoch, and provides a higher-precision forecast position for the current epoch, thereby extracting the comprehensive term of the prior error as a test quantity for coarse satellite screening.
[0069] The undifferenced Doppler observation equation is as follows:
[0070]
[0071] Where D is the Doppler observation value, r R and are the receiver position and velocity vectors, r S i and Satellite S i The position and velocity vectors, is the receiver clock change rate, is the satellite clock change rate, and are the delay change rates of the troposphere and ionosphere, respectively. Since the atmosphere changes slowly, the change rate can be ignored. γ is the comprehensive error term of Doppler observation noise and multipath error.
[0072] Linearize the above formula to get the function model of Doppler velocity measurement:
[0073]
[0074] Among them, m x 、m y 、m z represents the linearization coefficient of each satellite receiver pair, (VxVyVz) represents the instantaneous velocity of the carrier at the current epoch, is the receiver clock rate term, and the differences between each system and each frequency point can be ignored. If the position change of the carrier from the i-th to the i+1-th epoch is to be obtained, the constant acceleration model is often used, and the calculation is as follows:
[0075]
[0076] On this basis, the predicted position of the current epoch is obtained. By substituting the predicted position into the pseudorange double difference equation (6) and the phase double difference equation (7), the double difference residual of each satellite is extracted, and the outliers are eliminated using the Grubbs rule, thus completing the rough screening of satellites.
[0077] 3. Post-test data-driven star selection method.
[0078] In the GNSS positioning algorithm, the random model reflects the accuracy of different observation values. It is necessary to use a reasonable random model to reflect the statistical characteristics of GNSS observation data to the greatest extent, so as to effectively improve the positioning accuracy and performance of the receiver. Since the navigation module is affected by different multipath effects in different observation environments, the observation value characteristics are different, and it is difficult to rely on a certain empirical model to accurately reflect the size of random errors in the entire scene. Therefore, the present invention uses code phase combined double difference observations to estimate observation noise in real time to effectively solve the problem of poor applicability of empirical models. Construct the CMC combined observation value of satellite i:
[0079]
[0080] Among them, δt ρ and They are the receiver pseudorange clock error and carrier phase clock error respectively. The influence of the clock error cannot be eliminated by CMC combination, but can be weakened by inter-satellite difference. and They are pseudo-range multipath error and phase multipath error respectively. Since the maximum value of phase multipath error is 1 / 4 wavelength (about 5 cm), it can be ignored compared with the meter-level pseudo-range multipath error. i is the ambiguity term, which is a constant when the carrier phase is observed continuously. is the ionospheric delay error, which can be eliminated by inter-station difference in short baseline observation. Therefore, inter-station difference and inter-satellite difference are performed on the above formula to obtain the double-difference CMC combined observation value ( Figure 4 Ci in Chinese:
[0081]
[0082] Among them, ▽Δ is the double difference operator, and the double difference ambiguity is a constant term, which can be deducted by taking the triple difference with the previous epoch. Through post-analysis, it is found that the above double-difference CMC combined observations can reflect the comprehensive errors of pseudorange multipath and noise just like the double-difference pseudorange residuals extracted by post-inserting the true position, and then construct the variance-covariance matrix of the double-difference pseudorange observations. Since the comprehensive error terms such as carrier phase observation noise and multipath error are difficult to extract, the noise matrix W of the double-difference phase observations is obtained by the empirical ratio of pseudorange noise to phase noise.
[0083] First, determine whether the double-difference CMC combined observation value Ci has a cycle slip. If a cycle slip occurs, set C i is the initial error C 0 , observation noise W i W i-1 After that, construct the observation noise matrix W. Otherwise, construct the CMC triple difference observation C i -C i-1 , and form the observation noise matrix W.
[0084] Kalman filtering is performed based on the constructed observation noise matrix W. After Kalman filtering is completed, the standardized residual v of each phase observation value is calculated s,i , as shown below:
[0085]
[0086] Among them, v i is the posterior residual of phase observation i, σ i The diagonal elements of the variance matrix of the phase observations constructed by the CMC combined observations, if v s,i <k 0 ,k 0 =1.5, it is selected as the available floating-point ambiguity and enters the ambiguity fixing stage. Otherwise, after removing satellite i, Kalman filtering is performed again, the standardized residual is calculated, and it is judged again whether it passes the test, thereby completing the precise screening of satellites.
[0087] The present invention also relates to a GNSS RTK positioning and star selection system driven by both models and data, comprising:
[0088] The first module is configured to perform the first satellite elimination: using a hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio to eliminate satellites of poor quality;
[0089] The second module is configured to perform the second satellite elimination: pre-screening each remaining satellite according to its contribution to GDOP;
[0090] The third module is configured to perform the third satellite elimination: for the pre-screened satellites, the pre-verification pseudorange and phase double difference residuals are calculated based on their predicted positions at the current epoch, outliers are eliminated, and rough screening is completed.
[0091] The fourth module is configured to perform the fourth satellite elimination: the satellites after coarse screening enter the RTK floating point solution, and the satellites are finely screened based on the post-verification standardized residuals and then enter the ambiguity fixation.
[0092] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.
Claims
1. A GNSS RTK positioning and star selection method driven by both model and data, characterized by: include Perform the first satellite elimination: use the hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio to eliminate satellites with poor quality; Perform the second satellite elimination: pre-screen each remaining satellite based on its contribution to GDOP; Perform the third satellite elimination: for the pre-screened satellites, calculate their pre-verification pseudorange and phase double difference residuals based on their predicted positions at the current epoch, eliminate outliers, and complete the rough screening; The fourth satellite elimination is performed: the satellites after rough screening enter the RTK floating point solution, and the satellites are finely screened based on the post-verification standardized residuals and then enter the ambiguity fixation; Before the first satellite removal, the hybrid model of altitude angle and signal-to-noise ratio is constructed by collecting a large sample of GNSS raw observation data covering typical complex scenes in the city, substituting the true position value to extract the true pseudo-range noise of each satellite, and thus constructing a hybrid model of pseudo-range noise, altitude angle, and signal-to-noise ratio; By analyzing the correlation between pseudorange noise, signal-to-noise ratio and altitude angle, a hybrid model of signal-to-noise ratio and altitude angle is constructed to calculate the theoretical noise value; the theoretical noise model of satellite i at the fth frequency point is constructed as follows: Among them, q1, a1, b1, q2, a2, b2 are the empirical parameters to be fitted, el i represents the altitude angle of satellite i, SNR i,f represents the signal strength of the fth frequency point of satellite i. Substituting the real noise, signal-to-noise ratio and altitude angle extracted from large sample data into the above formula, six empirical parameters can be fitted; When real-time data is obtained, the theoretical noise model δ can be calculated by formula (1): i,f , when the theoretical noise satisfies δ i,f ≤δ0, it is selected as an available satellite.
2. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: When the third satellite elimination is performed, the Doppler velocity measurement solution is performed using the roughly screened satellites to obtain high-precision velocity information, and the carrier position change equation between epochs is constructed to calculate the predicted position of the current epoch. The predicted position is substituted into the pseudorange double difference and phase double difference observation equations to calculate the pre-test residuals, and the Grubbs criterion is used to eliminate outliers in the pre-test residuals to complete the rough screening of satellites.
3. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: When the fourth satellite is removed, the code phase combination double difference observation value is constructed to estimate the observation noise in real time, and then the variance covariance matrix of the double difference pseudorange / phase is constructed. After the floating point solution is completed, the posterior residual and the observation variance are used to calculate the posterior standardized residual v of each observation value. After removing the satellite with v exceeding the limit, the ambiguity fixation is entered.
4. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: Extract the pseudo-range noise of large sample original observation data in the following way: The above formula is the pseudo-range non-difference observation equation, where: is the original pseudorange observation value, ρ represents the satellite-to-earth distance, δt R and δt S Represent the receiver clock error and satellite clock error, δ trop and δ ion represent the tropospheric delay error and the ionospheric delay error, δ mul and ε ρ are the multipath error and pseudorange observation noise, respectively.
5. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: Calculate the contribution of each remaining satellite to GDOP The formula is as follows: Among them, h i is the observation vector corresponding to the i-th satellite, H n is the observation matrix corresponding to n satellites.
6. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: It is defined that observation stations u and r observe GNSS satellites i and j at the same time, then the GNSS pseudorange double difference observation equation and carrier phase double difference observation equation are as follows: Where P is the pseudorange observation value, λ is the carrier wavelength, is the carrier phase observation value, ρ is the geometric distance between the satellite and the receiver, I and T are the residual errors of ionospheric delay and tropospheric delay respectively. In the short baseline mode, the atmospheric residual error can be ignored, ε is the comprehensive term of pseudorange observation noise and multipath error, and ζ is the comprehensive term of phase observation noise and multipath error.
7. The GNSS RTK positioning and star selection method driven by both model and data as claimed in claim 1, characterized in that: Construct the CMC combined observation value of satellite i and perform inter-station difference and inter-satellite difference on it to obtain the double-difference CMC combined observation value: Among them, ▽Δ is the double difference operator, and the double difference ambiguity is a constant term, and the variance matrix of the double difference phase observation value is obtained by the empirical ratio of pseudorange noise to phase noise; And determine whether a cycle slip occurs. If a cycle slip occurs, let C i is the initial error C0, the observation noise W i W i-1 After that, construct the observation noise matrix W, otherwise, construct the CMC triple difference observation C i -C i-1 , and form the observation noise matrix W; Kalman filtering is performed based on the constructed observation noise matrix W. After Kalman filtering is completed, the standardized residual v of each phase observation value is calculated s,i , as shown below: Among them, v i is the posterior residual of phase observation i, σ i The diagonal elements of the variance matrix of the phase observations constructed by the CMC combined observations, if v s,i If k0 is less than k0, k0=1.5, the available floating-point ambiguity is selected and the ambiguity fixing stage is entered. Otherwise, the Kalman filter is performed again after satellite i is removed, the standardized residual is calculated, and it is judged again whether it passes the test.
8. A GNSS RTK positioning and star selection system driven by both models and data, characterized by: include The first module is configured to perform the first satellite elimination: using a hybrid model of pseudorange noise, altitude angle, and signal-to-noise ratio to eliminate satellites of poor quality; The second module is configured to perform the second satellite elimination: pre-screening each remaining satellite according to its contribution to GDOP; The third module is configured to perform the third satellite elimination: for the pre-screened satellites, the pre-verification pseudorange and phase double difference residuals are calculated according to the predicted positions of the current epochs, and outliers are eliminated to complete the rough screening; The fourth module is configured to perform the fourth satellite elimination: the satellites after the rough screening enter the RTK floating point solution, and the satellites are finely screened based on the post-verification standardized residual and then enter the ambiguity fixation; Before the first satellite removal, the hybrid model of altitude angle and signal-to-noise ratio is constructed by collecting a large sample of GNSS raw observation data covering typical complex scenes in the city, substituting the true position value to extract the true pseudo-range noise of each satellite, and thus constructing a hybrid model of pseudo-range noise, altitude angle, and signal-to-noise ratio; By analyzing the correlation between pseudorange noise, signal-to-noise ratio and altitude angle, a hybrid model of signal-to-noise ratio and altitude angle is constructed to calculate the theoretical noise value; the theoretical noise model of satellite i at the fth frequency point is constructed as follows: Among them, q1, a1, b1, q2, a2, b2 are the empirical parameters to be fitted, el i represents the altitude angle of satellite i, SNR i,f represents the signal strength of the fth frequency point of satellite i. Substituting the real noise, signal-to-noise ratio and altitude angle extracted from large sample data into the above formula, six empirical parameters can be fitted; When real-time data is obtained, the theoretical noise δ can be calculated by formula (2): i,f , when the theoretical noise satisfies δ i,f ≤δ0, it is selected as an available satellite.