Real-time dynamic positioning integer ambiguity resolving method and system based on BIE

Abstract

The invention relates to the technical field of GNSS high-precision positioning, and provides a BIE-based real-time dynamic positioning integer ambiguity resolving method and system, and the method comprises the steps: obtaining a preliminary ambiguity subset through the screening of a satellite elevation angle and a carrier-to-noise ratio, and carrying out the correlation reduction processing; a multi-level screening mechanism is adopted and comprises three stages of dynamic optimization of coarse screening, fine screening and candidate group generation; a weight distribution model is improved based on Laplacian distribution to calculate candidate solution weights and probability weighted fusion is carried out; in combination with an ILS and BIE mixed fixing strategy, BIE weighted averaging is adopted when ILS fixing fails; a positioning result is optimized through dynamic satellite screening and multi-system data fusion, and a jump detection threshold value is set to realize abnormal recalculation. The method improves the reliability of real-time dynamic positioning, and guarantees the continuous and stable operation of the system in a signal shielding or multi-path interference scene through a jump detection and recalculation mechanism.

Description

Technical Field

[0001] This invention relates to the field of GNSS high-precision positioning technology, and in particular to a method and system for real-time dynamic positioning integer ambiguity resolution based on BIE. Background Technology

[0002] GNSS (Global Navigation Satellite System) RTK (Real-Time Kinematic) technology is a key technology for achieving centimeter-level high-precision positioning through carrier phase observation. It is widely used in critical application areas such as autonomous driving, surveying and mapping, precision agriculture, and drone navigation. In practical applications, RTK technology needs to solve the core technical challenge of integer ambiguity, namely, determining the number of integer cycles contained in the carrier phase observation values. The accuracy and efficiency of integer ambiguity resolution directly affect the accuracy and reliability of RTK positioning. However, low-cost GNSS receivers face numerous technical challenges in fixing integer ambiguity in complex environments, including high observation noise, severe multipath interference, and low ambiguity fixing success rates. The key lies in effectively dealing with complex electromagnetic interference, accurately identifying high-quality ambiguity candidate solutions, and achieving highly reliable real-time ambiguity fixing.

[0003] In existing technologies, integer ambiguity resolution mainly employs the traditional ILS (Indirect Least Squares) method and the basic BIE (best integer equivariant) method, achieving basic ambiguity fixation. However, existing methods do not adequately consider the inherent statistical correlation mechanism between the complex observation environment of low-cost GNSS receivers and the accuracy of ambiguity resolution. They struggle to organically integrate the objective characteristics of observation noise distribution with the actual complex urban environment, resulting in the inability to achieve high-precision, high-reliability, low-cost automatic integer ambiguity resolution. Summary of the Invention

[0004] In view of this, the present invention proposes a real-time dynamic positioning integer ambiguity resolution method and system based on BIE, which solves the problem that existing methods do not adequately consider the inherent statistical correlation mechanism between the complex observation environment of low-cost GNSS receivers and the accuracy of ambiguity resolution, and are difficult to organically integrate the objective observation noise distribution characteristics with the actual complex urban environment characteristics, resulting in the inability to achieve high-precision, high-reliability, low-cost automatic RTK integer ambiguity resolution.

[0005] The technical solution of this invention is implemented as follows: On one hand, this invention provides a real-time dynamic positioning integer ambiguity resolution method based on BIE, including the following steps: The GNSS observation data is filtered, and abnormal satellites in the GNSS observation data are removed according to the satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset; The pre-selected fuzzy subset is subjected to decorrelation processing to obtain a decorrelated fuzzy subset; A multi-level filtering mechanism is used to dynamically optimize the decorrelated fuzziness subset to obtain a candidate group of fuzziness. Based on the weight allocation model improved by Laplace distribution, the weights of each candidate solution in the candidate group of fuzziness screening are calculated, and the candidate group of fuzziness screening is probabilistically weighted and fused to obtain the fused fuzziness. The fused fuzziness is fixed by the ILS method. When the ILS fixation fails, the fused fuzziness is weighted and averaged by the BIE method to obtain the fixed fuzziness. The initial positioning result is calculated based on the fixed ambiguity. The initial positioning result is optimized by a dynamic satellite filtering strategy to obtain an optimized positioning result. The error influence of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

[0006] Based on the above technical solutions, preferably, the multi-level screening mechanism includes a coarse screening stage, a fine screening stage, and a candidate group generation stage, wherein... The coarse screening stage includes: removing satellites with a PDOP value higher than 3.0 based on the satellite geometry configuration PDOP value, and removing satellites with a carrier-to-noise ratio lower than 35dBHz based on the signal-to-noise ratio, to obtain a coarsely screened subset of decorrelated ambiguity. The fine screening stage includes: evaluating the bootstrapping success rate of the decorrelated ambiguity subset after coarse screening, calculating the fixed probability of each ambiguity, screening ambiguities with a fixed probability higher than 90%, and simultaneously deleting low-quality ambiguities in descending order of Z-transform decorrelated variance, dynamically adjusting the size of the ambiguity subset to 6-10, and obtaining the finely screened ambiguity subset. The candidate group generation stage includes: setting the maximum number of candidate groups to 100 based on the finely filtered fuzzy subset, calculating the weight of each candidate solution using a weight allocation model improved by Laplace distribution, and obtaining the filtered fuzzy candidate groups.

[0007] Based on the above technical solutions, preferably, the Bootstrapping success rate assessment in the fine screening stage is based on calculating the fixed probability of ambiguity using conditional variance, and high fixed probability ambiguities are screened by evaluating the strength of the GNSS mathematical model. The Z-transform decorrelation variance screening sorts the ambiguity variance-covariance matrix by decorrelation sorting, prioritizing the retention of high-quality ambiguities with small variances and deleting low-quality ambiguities with large variances. When the size of the ambiguity subset exceeds 10, the ambiguities are deleted from largest to smallest according to the decreasing correlation variance and the ADOP value is recalculated to ensure that the ADOP value is less than 0.0055. When the size of the ambiguity subset is less than 6, GNSS observation data is re-acquired and the ambiguities corresponding to the satellites that meet the conditions are supplemented and filtered. During the candidate group generation phase, the Laplace distribution improved weight allocation model replaces the exponentially decaying weights of the traditional Gaussian distribution with linearly decaying weights.

[0008] Based on the above technical solutions, preferably, the probability weighted fusion includes an ILS fixed judgment stage and a BIE weighted average stage; The ILS fixing judgment stage includes: fixing the ambiguity of the selected ambiguity candidate group using the ILS method, and determining whether the ILS fixing is successful by combining the Ratio test and the F test. When the Ratio test threshold is greater than 3.0 and the significance level of the F test is greater than 0.05, the ILS fixing result is output to obtain the fixed ambiguity. The BIE weighted averaging stage includes: when ILS fixation fails, calculating the likelihood function weight of the fused ambiguity based on the Laplace distribution probability density function, replacing the traditional Gaussian distribution exponential decay model with a linear decay weight model, and performing a weighted average on all candidate solutions of the fused ambiguity to obtain the fixed ambiguity after BIE weighted averaging. Fixing the fused ambiguities includes: combining the PAR-BIE strategy, partial ambiguity fixing, and optimal integer equivariant estimation; fixing satellite combinations with PDOP values ​​less than 2.0 and Bootstrapping fixing probability greater than 95%; when the size of the ambiguity subset exceeds 10, deleting ambiguities from largest to smallest according to the Z-transform to reduce correlation variance, and recalculating the ADOP value; when the size of the ambiguity subset is less than 6, re-collecting GNSS observation data and supplementing the ambiguities corresponding to the satellites.

[0009] Based on the above technical solution, preferably, the scaling factor value of the Laplace distribution probability density function in the BIE weighted average stage is set to 4, the geometric distance between each candidate solution and the floating-point solution is calculated based on the scaling factor, and the geometric distance is converted into a weight value through the Laplace distribution function. The linear decay weight model adopts the double exponential decay characteristic of the Laplace distribution, assigning relatively higher weights to candidate solutions that are far from the floating-point solution and relatively lower weights to candidate solutions that are close to the floating-point solution. By calculating the PDOP value and Bootstrapping fixation probability of each satellite combination, satellite combinations with a PDOP value less than 2.0 and a Bootstrapping fixation probability greater than 95% are selected for priority fixation. When the size of the ambiguity subset exceeds 10, ambiguity parameters are deleted one by one in descending order of the Z-transform downcorrelation variance value in the ambiguity variance-covariance matrix. After deletion, the ADOP value is recalculated and compared with the second preset threshold. When the ADOP value is less than the second preset threshold, the current ambiguity subset is retained. When the size of the ambiguity subset is less than 6, GNSS observation data is reacquired and satellites are added.

[0010] Based on the above technical solutions, preferably, the initial positioning result is calculated based on the fixed ambiguity, and the initial positioning result is optimized through a dynamic satellite filtering strategy to obtain an optimized positioning result. Then, the error impact of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is re-performed, including: Position is calculated using the carrier phase observation equation. The PDOP threshold is set to 2.0 and the carrier-to-noise ratio threshold is set to 35 dBHz. Satellites exceeding the threshold range are removed to obtain a filtered satellite set. The optimized positioning result is obtained by recalculating based on the filtered satellite set. The multi-system data fusion is achieved by fusing observation data from GPS, BDS, and Galileo systems using a weighted least squares method. The weighting coefficients of the GPS, BDS, and Galileo system observation data are set according to the observation noise variance of each system to correct the optimized positioning result and obtain the final positioning result. The first preset threshold is set to 0.05m. When the absolute value of the coordinate difference between the final positioning result and the previous epoch positioning result exceeds the first preset threshold, the calculation is performed again.

[0011] Based on the above technical solutions, preferably, the carrier phase observation equation is based on a double-difference observation model to eliminate receiver clock bias and satellite clock bias; The filtered satellite set is determined by satellite continuity between epochs by recording the PDOP value and carrier-to-noise ratio data of satellites in two consecutive epochs. When a satellite continuously meets the conditions of PDOP value less than 2.0 and carrier-to-noise ratio greater than 35dBHz, the satellite is included in the filtered satellite set. The weighted least squares method includes the following weighting coefficients for the GPS system: the pseudorange noise variance system weighting coefficient is 0.3, the carrier phase noise variance system weighting coefficient is 0.001, and the weighting coefficients for the BDS and Galileo systems are adjusted proportionally according to the signal quality. The calculation of the absolute value of the coordinate difference includes the coordinate component differences in the three directions of E, N, and U; The recalculation process includes re-filtering GNSS observation data, re-performing decorrelation processing, re-filtering at multiple levels, and re-performing BIE probability weighted fusion until the final positioning result meets the accuracy requirements.

[0012] Based on the above technical solutions, preferably, the step of filtering GNSS observation data by removing abnormal satellites from the GNSS observation data according to satellite elevation angle, carrier-to-noise ratio, and observation residuals, to obtain a preliminary ambiguity subset, includes: Satellites with an elevation angle below 15° and a carrier-to-noise ratio below 35dBHz were removed to obtain the observation data after filtering by elevation angle and carrier-to-noise ratio; The observation data after being filtered by elevation angle and carrier-to-noise ratio are subjected to robust adaptive Kalman filtering, the observation residuals of each satellite are calculated, and satellites with observation residuals exceeding twice the root mean square error are removed to obtain the preliminary ambiguity subset.

[0013] Based on the above technical solutions, preferably, the step of performing decorrelation processing on the initially screened fuzzy subset to obtain a decorrelated fuzzy subset includes: The initially screened ambiguity subset is subjected to Z-transform decorrelation processing, and the ambiguity variance-covariance matrix is ​​calculated using the LAMBDA method to obtain the ambiguity variance-covariance matrix after decorrelation processing. The ADOP value is calculated based on the ambiguity variance-covariance matrix after the downcorrelation processing. The ADOP value is compared with a second preset threshold. When the ADOP value is lower than the second preset threshold, the downcorrelation ambiguity subset is obtained. When the ADOP value is higher than the second preset threshold, the ambiguity subset is re-optimized.

[0014] On the other hand, the present invention also provides a real-time dynamic positioning integer ambiguity resolution system based on BIE, the system comprising: The GNSS observation data filtering module is used to filter GNSS observation data. It removes abnormal satellites from the GNSS observation data based on satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset. The fuzziness decorrelation processing module is used to perform decorrelation processing on the initially screened fuzziness subset to obtain a decorrelated fuzziness subset. A multi-level fuzziness filtering and optimization module is used to dynamically optimize the decorrelated fuzziness subset using a multi-level filtering mechanism to obtain a fuzziness candidate group; The fusion weight calculation and ambiguity fixing module is used to calculate the weight of each candidate solution in the ambiguity candidate group based on the weight allocation model improved by Laplace distribution, perform probability weighted fusion on the ambiguity candidate group to obtain fused ambiguity, fix the fused ambiguity using the ILS method, and when the ILS fixing fails, the BIE method is used to perform weighted averaging on the fused ambiguity to obtain fixed ambiguity. The positioning result calculation and multi-system fusion correction module is used to calculate the initial positioning result based on the fixed ambiguity, optimize the initial positioning result through a dynamic satellite filtering strategy to obtain an optimized positioning result, and combine multi-system data fusion to correct the error influence of the optimized positioning result to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

[0015] The real-time dynamic positioning integer ambiguity resolution method and system based on BIE of the present invention has the following advantages over the prior art: (1) By combining GNSS multi-dimensional screening preprocessing and Laplace distribution weight allocation model, a multi-level dynamic screening mechanism is used to generate high-quality ambiguity candidate groups and BIE probability weighted fusion. The optimal solution path is dynamically selected by combining ILS and BIE hybrid fixed strategy. The positioning results are optimized and corrected in real time by multi-system data fusion and dynamic satellite screening, which improves the reliability of real-time dynamic positioning. At the same time, the jump detection and recalculation mechanism ensures the continuous and stable operation of the system in the scenario of signal blockage or multi-path interference. (2) By constructing a three-level progressive screening architecture of coarse screening, fine screening and candidate group generation, the initial quality control is carried out by the dual constraints of PDOP geometric configuration and carrier-to-noise ratio. The precise screening is carried out by combining Bootstrapping success rate evaluation and Z-transform correlation reduction variance sorting. The optimal solution conditions are maintained by dynamically adjusting ADOP value and adaptively controlling the size of ambiguity subset, thereby improving the quality of ambiguity candidate groups in complex environments. (3) By constructing a dual guarantee mechanism of ILS fixed judgment and BIE weighted average, the reliability of ILS fixed is ensured by using the Ratio test and F test to jointly determine the reliability of ILS fixed. The traditional Gaussian exponential decay model is replaced by the Laplace distribution probability density function and the linear decay weight model with a scaling factor of 4. The PAR-BIE strategy is used to prioritize the fixed processing of high-quality satellite combinations, which improves the success rate of integer ambiguity resolution and the stability of positioning solution. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a real-time dynamic positioning integer ambiguity resolution method based on BIE according to the present invention; Figure 2 This is a schematic diagram of the PAR-BIE ambiguity fixing scheme of the present invention; Figure 3 This is a schematic diagram comparing the positioning error sequences of different algorithms in complex environments according to embodiments of the present invention; Figure 4 This is a schematic diagram comparing the convergence time of the algorithm in different scenarios according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This invention provides a real-time dynamic positioning integer ambiguity resolution method based on BIE, comprising the following steps: The GNSS observation data is filtered, and abnormal satellites in the GNSS observation data are removed according to the satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset; The pre-selected fuzzy subset is subjected to decorrelation processing to obtain a decorrelated fuzzy subset; A multi-level filtering mechanism is used to dynamically optimize the decorrelated fuzziness subset to obtain a candidate group of fuzziness. Based on the weight allocation model improved by Laplace distribution, the weights of each candidate solution in the candidate group of fuzziness screening are calculated, and the candidate group of fuzziness screening is probabilistically weighted and fused to obtain the fused fuzziness. The fused fuzziness is fixed by the ILS method. When the ILS fixation fails, the fused fuzziness is weighted and averaged by the BIE method to obtain the fixed fuzziness. The initial positioning result is calculated based on the fixed ambiguity. The initial positioning result is optimized by a dynamic satellite filtering strategy to obtain an optimized positioning result. The error influence of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

[0020] Specifically, this embodiment addresses the problem of erroneous fixation in traditional integer least squares under observation noise interference by using an improved BIE method based on the Laplace distribution. It changes the exponentially decaying weights of the Gaussian distribution to a linearly decaying form, mitigating the risk of weight concentration in a single candidate solution. Especially when satellite signal obstruction leads to variance expansion of floating-point solutions, it assigns non-zero weights to multiple candidate solutions, avoiding positioning jumps caused by erroneous fixation. It also incorporates PAR-BIE (Partial Ambiguity Resolution-Best Integer Equivariant)... Estimation (partial ambiguity resolution - best integer equivariant estimation) strategy, partial ambiguity fixation and BIE optimization strategy, dynamically optimizes the ambiguity subset through a multi-level screening mechanism, and combines probabilistic weighted fusion of candidate solutions to significantly improve ambiguity fixation rate and convergence efficiency. Abnormal satellites are removed based on satellite elevation angle, carrier-to-noise ratio, and observation residuals, retaining reliable observation data. Ambiguity variance-covariance information and reduced correlation ranking are used. The ambiguity subset is dynamically optimized through Bootstrapping success rate assessment and Z-transform reduced correlation variance screening. When ILS fixation fails, the BIE method is used to probabilistically weight all potential candidate solutions, integrating information from multiple candidate solutions to reduce the impact of multipath and low-quality observations on the positioning results. By setting a maximum number of candidate groups (e.g., 100 groups) and strict threshold conditions, such as ADOP (Ambiguity Dilution of...),... Precision (ambiguity accuracy impact factor) value of 0.0055 balances computational complexity and solution accuracy. At the same time, a dynamic satellite selection strategy is introduced to prioritize the retention of satellite combinations with strong geometric configurations in complex scenarios such as tall buildings blocking the view or under overpasses, thereby improving positioning robustness.

[0021] This embodiment combines GNSS multi-dimensional screening preprocessing with a Laplace distribution weight allocation model, utilizes a multi-level dynamic screening mechanism to generate high-quality ambiguity candidate groups and perform BIE probability weighted fusion, dynamically selects the optimal solution path using a hybrid ILS and BIE fixed strategy, and optimizes and corrects the positioning results in real time through multi-system data fusion and dynamic satellite screening. This solves the problems of low success rate of ambiguity fixation and unstable positioning accuracy of low-cost GNSS receivers in complex environments, improves the reliability of real-time dynamic positioning, and ensures the continuous and stable operation of the system in scenarios with signal obstruction or multi-path interference through jump detection and recalculation mechanisms.

[0022] This invention not only significantly improves the positioning accuracy and stability of low-cost RTK systems, but also provides an efficient and low-cost solution for high-precision dynamic positioning in complex urban environments through theoretical innovation and engineering optimization, and has broad application prospects.

[0023] The process of filtering GNSS observation data involves removing abnormal satellites based on satellite elevation angle, carrier-to-noise ratio, and observation residuals to obtain a preliminary ambiguity subset, including: Satellites with an elevation angle below 15° and a carrier-to-noise ratio below 35dBHz were removed to obtain the observation data after filtering by elevation angle and carrier-to-noise ratio.

[0024] In one specific embodiment, the satellite elevation angle-based filtering employs an adaptive threshold strategy, including: During the floating-point solution stage, a 10° elevation angle threshold is used to screen satellites to obtain more observation data. During the ambiguity fixing stage, a 15° elevation angle threshold is used to screen satellites to ensure observation quality and eliminate low elevation angle satellites that are severely affected by multipath and tropospheric errors. The carrier-to-noise ratio-based screening uses 35dBHz as a uniform threshold to eliminate satellite observation data with poor signal quality. The synergistic effect of the dual screening criteria ensures the reliability of the data foundation for subsequent ambiguity resolution.

[0025] The observation data after being filtered by elevation angle and carrier-to-noise ratio are subjected to robust adaptive Kalman filtering, the observation residuals of each satellite are calculated, and satellites with observation residuals exceeding twice the root mean square error are removed to obtain the preliminary ambiguity subset.

[0026] In one specific embodiment, the robust adaptive Kalman filtering process suppresses the influence of abnormal observations by dynamically adjusting the filter gain matrix and calculates the residuals of carrier phase observations for each satellite. By performing residual comparison analysis on observations of the same satellite at different frequencies, gross errors are detected through geometrically independent combination of dual-frequency observations. When the residual of the same-frequency observation exceeds twice the root mean square error or the residual of the different-frequency observation is significantly different, the corresponding satellite is removed. This fine screening mechanism can effectively exclude observation data affected by gross errors, cycle slips and multipath effects, thereby providing a high-quality ambiguity subset basis.

[0027] The dual-frequency difference relationship of the fine screening mechanism is as follows: ; in, and Receiver With satellite The carrier phase observation, expressed in units of length, is multiplied by the wavelength, corresponding to the frequency points. and . and Corresponding frequency points and The carrier wavelength; and Receiver With satellite The integer ambiguity is represented by the frequency points. and . and Corresponding frequency points and The equivalent terms of geometry and correction terms in phase observations include geometric distance correction, tropospheric, clock error, and other terms merged according to specific models; Two frequency points and The combined observation noise after differencing and the unmodeled residual term.

[0028] The step of performing decorrelation processing on the initially screened fuzzy subset to obtain a decorrelated fuzzy subset includes: The initially screened ambiguity subset is subjected to Z-transform decorrelation processing, and the ambiguity variance-covariance matrix is ​​calculated using the LAMBDA method to obtain the ambiguity variance-covariance matrix after decorrelation processing.

[0029] In one specific embodiment, the Z-transform decorrelation processing obtains a lower triangular matrix and a diagonal matrix by performing Cholesky decomposition on the initially screened ambiguity subset, and then reduces the correlation between ambiguities by optimizing the matrix structure through decorrelation transformation. When calculating the ambiguity variance-covariance matrix using the LAMBDA method, the floating-point solution of the ambiguity is first decomposed by Cholesky, and then the ambiguity variance-covariance matrix is ​​decorated by an integer transformation matrix so that the diagonal elements of the ambiguity variance-covariance matrix after the decorrelation processing can more accurately reflect the precision information of a single ambiguity.

[0030] The ADOP value is calculated based on the ambiguity variance-covariance matrix after the downcorrelation processing. The ADOP value is compared with a second preset threshold. When the ADOP value is lower than the second preset threshold, the downcorrelation ambiguity subset is obtained. When the ADOP value is higher than the second preset threshold, the ambiguity subset is re-optimized.

[0031] In one specific embodiment, the ADOP value is calculated based on the diagonal elements of the ambiguity variance-covariance matrix after the downcorrelation processing, and the ambiguity solvability evaluation index is obtained by geometric mean. The second preset threshold is set to 0.0055. When the ADOP value is lower than this threshold, it indicates that the ambiguity has fixed conditions. When the ADOP value is higher than this threshold, it indicates that the quality of the ambiguity subset is insufficient to support reliable fixed solutions. This adaptive evaluation mechanism ensures a good data foundation for subsequent BIE probability-weighted fusion by dynamically adjusting the ambiguity subset, avoiding complex candidate group screening and weight calculation on low-quality ambiguity subsets, thereby improving overall solution efficiency and reducing the risk of error fixation.

[0032] Please see Figure 2 This is a schematic diagram of the PAR-BIE ambiguity fixing scheme of the present invention. The multi-level screening mechanism includes a coarse screening stage, a fine screening stage, and a candidate group generation stage. The coarse screening stage includes: removing satellites with a PDOP value higher than 3.0 based on the satellite geometry configuration PDOP (Position Dilution of Precision) value, and removing satellites with a carrier-to-noise ratio lower than 35dBHz based on the signal-to-noise ratio, thus obtaining a coarsely screened subset of decorrelated ambiguity. The fine screening stage includes: evaluating the bootstrapping success rate of the decorrelated ambiguity subset after coarse screening, calculating the fixed probability of each ambiguity, screening ambiguities with a fixed probability higher than 90%, and simultaneously deleting low-quality ambiguities in descending order of Z-transform decorrelated variance, dynamically adjusting the size of the ambiguity subset to 6-10, and obtaining the finely screened ambiguity subset. The candidate group generation stage includes: setting the maximum number of candidate groups to 100 based on the finely filtered fuzzy subset, calculating the weight of each candidate solution using a weight allocation model improved by Laplace distribution, and obtaining the filtered fuzzy candidate groups.

[0033] The Bootstrapping success rate assessment in the fine screening stage is based on the conditional variance calculation of the fixed probability of ambiguity, and high fixed probability ambiguities are screened by evaluating the strength of the GNSS mathematical model. The Z-transform decorrelation variance screening sorts the ambiguity variance-covariance matrix by decorrelation sorting, prioritizing the retention of high-quality ambiguities with small variances and deleting low-quality ambiguities with large variances. When the size of the ambiguity subset exceeds 10, the ambiguities are deleted from largest to smallest according to the decreasing correlation variance and the ADOP value is recalculated to ensure that the ADOP value is less than 0.0055. When the size of the ambiguity subset is less than 6, GNSS observation data is re-acquired and the ambiguities corresponding to the satellites that meet the conditions are supplemented and filtered. During the candidate group generation stage, the Laplace distribution improved weight allocation model replaces the exponential decay weight of the traditional Gaussian distribution with linear decay weights to avoid excessive concentration of weights in a single candidate solution and reduce the risk of incorrect fixation in the heavy-tailed error scenario.

[0034] In one specific embodiment, the formula for calculating the success rate of Bootstrapping is: ; in, Bootstrapping success rate is a joint probability index representing the successful fixation of ambiguity. The number of ambiguity parameters; For the first The ambiguity is in front The conditional standard deviation under a fixed ambiguity is obtained by the ambiguity variance-covariance matrix and a fixed order. This is the cumulative distribution function of the standard normal distribution.

[0035] The cumulative distribution function of the standard normal distribution is calculated as follows: ; in, and As the independent variable, It is an exponential function.

[0036] The parameter estimation formula for LBIE (Laplace best integer equivariant, a weighted distribution model improved by Laplace distribution) is as follows: ; ; ; in, The ambiguity estimate obtained from the LBIE model is the fused and weighted estimate. Estimate the baseline parameters obtained from the LBIE model; A candidate ambiguity vector at integer grid points. ; For the first Candidates The weights are constructed based on the likelihood of the Laplace distribution; The variance-covariance submatrix of the ambiguity parameters; This is the covariance submatrix between the baseline and the ambiguity. and These are the floating-point solutions for ambiguity and the floating-point estimates for baseline parameters, respectively. for 1-dimensional integer lattice; The variance-covariance matrix of the floating-point solution for ambiguity; It is an exponential function. This embodiment uses linearly decaying weights instead of the exponentially decaying weights of the traditional Gaussian BIE to alleviate the problem of weight concentration in a single candidate solution.

[0037] This embodiment constructs a three-level progressive screening architecture: coarse screening, fine screening, and candidate group generation. It uses the dual constraints of PDOP geometry and carrier-to-noise ratio for initial quality control, and combines Bootstrapping success rate evaluation and Z-transform-based correlation variance sorting for precise screening. It maintains optimal solution conditions through dynamic adjustment of ADOP value and adaptive control of ambiguity subset size. It adopts Laplace distributed linear decay weights instead of traditional Gaussian exponential decay weights to solve the problem of excessive weight concentration in heavy-tailed error scenarios, thereby improving the quality of ambiguity candidate groups in complex environments. At the same time, the multi-level quality threshold and dynamic size adjustment mechanism ensure the stability and convergence efficiency of the algorithm under different observation conditions.

[0038] The probability-weighted fusion includes an ILS fixed judgment phase and a BIE weighted average phase; The ILS fixing judgment stage includes: fixing the ambiguity of the selected ambiguity candidate group using the ILS method, and determining whether the ILS fixing is successful by combining the Ratio test and the F test. When the Ratio test threshold is greater than 3.0 and the significance level of the F test is greater than 0.05, the ILS fixing result is output to obtain the fixed ambiguity. The BIE weighted averaging stage includes: when ILS fixation fails, calculating the likelihood function weight of the fused ambiguity based on the Laplace distribution probability density function, replacing the traditional Gaussian distribution exponential decay model with a linear decay weight model, and performing a weighted average on all candidate solutions of the fused ambiguity to obtain the fixed ambiguity after BIE weighted averaging. Fixing the fused ambiguities includes: using the PAR-BIE strategy, fixing satellite combinations with PDOP values ​​less than 2.0 and Bootstrapping fixation probability greater than 95%; when the size of the ambiguity subset exceeds 10, deleting ambiguities from largest to smallest by reducing correlation variance using Z-transform, and recalculating the ADOP value; when the size of the ambiguity subset is less than 6, re-collecting GNSS observation data and supplementing the ambiguities corresponding to the satellites.

[0039] In the BIE weighted average stage, the scaling factor of the probability density function of the Laplace distribution is set to 4. Based on the scaling factor, the geometric distance between each candidate solution and the floating-point solution is calculated, and the geometric distance is converted into a weight value through the Laplace distribution function. The linear decay weight model adopts the double exponential decay characteristic of the Laplace distribution, assigning relatively higher weights to candidate solutions that are far from the floating-point solution and relatively lower weights to candidate solutions that are close to the floating-point solution. By calculating the PDOP value and Bootstrapping fixation probability of each satellite combination, satellite combinations with a PDOP value less than 2.0 and a Bootstrapping fixation probability greater than 95% are selected for priority fixation. When the size of the ambiguity subset exceeds 10, ambiguity parameters are deleted one by one in descending order of the Z-transform downcorrelation variance value in the ambiguity variance-covariance matrix. After deletion, the ADOP value is recalculated and compared with the second preset threshold. When the ADOP value is less than the second preset threshold, the current ambiguity subset is retained. When the size of the ambiguity subset is less than 6, GNSS observation data is reacquired and satellites are added.

[0040] In one specific embodiment, the Laplace probability density function is calculated as follows: ; in, Let Laplace's probability density function be . This is a vector of GNSS observations; and Design matrices for the observation models respectively, and The coefficient matrix corresponding to the integer ambiguity vector, The coefficient matrix corresponding to the baseline vector; This is the integer ambiguity vector; The baseline vector represents the location-dependent real-valued parameters. (or ) is the observation vector The variance-covariance matrix; Represents the determinant operation of a matrix; For The norm of the quadratic form with weights; This is the Laplace distribution scaling factor, which is set to 4 according to the document. It is an exponential function.

[0041] This embodiment constructs a dual guarantee mechanism of ILS fixing judgment and BIE weighted averaging. It uses the Ratio test and F test to jointly determine the reliability of ILS fixing. It replaces the traditional Gaussian exponential decay model with a Laplace distribution probability density function and a linear decay weight model with a scaling factor of 4. It adopts the PAR-BIE strategy to prioritize fixing high-quality satellite combinations. It solves the problems of high failure rate of ambiguity fixing and uneven weight distribution in complex environments by using the dual constraints of PDOP value and Bootstrapping fixing probability and dynamic adjustment of ADOP value. This improves the success rate of integer ambiguity resolution and the stability of positioning solution.

[0042] The initial positioning result is calculated based on the fixed ambiguity. The initial positioning result is then optimized using a dynamic satellite filtering strategy to obtain an optimized positioning result. Finally, the error impact of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is repeated, including: Position is calculated using the carrier phase observation equation. The PDOP threshold is set to 2.0 and the carrier-to-noise ratio threshold is set to 35 dBHz. Satellites exceeding the threshold range are removed to obtain a filtered satellite set. The optimized positioning result is obtained by recalculating based on the filtered satellite set. The multi-system data fusion is achieved by fusing observation data from GPS, BDS, and Galileo systems using a weighted least squares method. The weighting coefficients of the GPS, BDS, and Galileo system observation data are set according to the observation noise variance of each system to correct the optimized positioning result and obtain the final positioning result. The first preset threshold is set to 0.05m. When the absolute value of the coordinate difference between the final positioning result and the previous epoch positioning result exceeds the first preset threshold, the calculation is performed again.

[0043] The carrier phase observation equation is based on a double-difference observation model to eliminate receiver clock bias and satellite clock bias. The filtered satellite set is determined by satellite continuity between epochs by recording the PDOP value and carrier-to-noise ratio data of satellites in two consecutive epochs. When a satellite continuously meets the conditions of PDOP value less than 2.0 and carrier-to-noise ratio greater than 35dBHz, the satellite is included in the filtered satellite set. The weighted least squares method includes the following weighting coefficients for the GPS system: the pseudorange noise variance system weighting coefficient is 0.3, the carrier phase noise variance system weighting coefficient is 0.001, and the weighting coefficients for the BDS and Galileo systems are adjusted proportionally according to the signal quality. The calculation of the absolute value of the coordinate difference includes the coordinate component differences in the three directions of E, N, and U; The recalculation process includes re-filtering GNSS observation data, re-performing decorrelation processing, re-filtering at multiple levels, and re-performing BIE probability weighted fusion until the final positioning result meets the accuracy requirements.

[0044] In one specific embodiment, the GPS system observation data includes L1 / L5 frequency points, the BDS system observation data includes B1I / B2a frequency points, and the Galileo system observation data includes E1 / E5a ​​frequency points. Ionospheric delay is eliminated by ionospheric-free combination, and weighted least squares method is used for joint adjustment. The weight coefficients are determined by the variance of the observation noise of each system. For example, the pseudorange noise variance of GPS L1 frequency point is 0.3m², and the carrier phase noise variance is 0.001m². The positioning jump value is the absolute value of the difference between adjacent epoch coordinates. When the jump values ​​of three or more consecutive epochs exceed the threshold, a reset is triggered, and the calculation is performed again.

[0045] The core performance indicators of the statistical positioning results include ambiguity fixation rate, convergence time, and RMS values ​​in the E / N / U directions. The ambiguity fixation rate is obtained by statistically analyzing the proportion of successfully fixed ambiguities across all epochs, with a preset standard of ≥90%. The convergence time is calculated by statistically analyzing the time from signal recovery to the first fixation of ambiguity, with a preset standard of ≤30s. The preset standards for the RMS values ​​in the E / N / U directions are E≤0.1m, N≤0.1m, and U≤0.15m, respectively. If any indicator fails to meet the preset standard, the observation data under scenarios with poor signal quality and poor geometric configuration are fed back to the data filtering stage to optimize the filtering threshold. The carrier-to-noise ratio threshold can be adjusted according to the signal interference intensity. Threshold optimization further improves the model's adaptability to different complex environments. The data feedback cycle is set to once every 24 hours, and batch optimization is performed based on the statistical results of daily observation data to ensure that the model performance continues to improve with the accumulation of application scenarios.

[0046] In one specific embodiment, observation data is used to train and verify the performance of the integer ambiguity resolution method, which is an important prerequisite for carrying out low-cost RTK high-precision positioning. The quantity and quality of observation data play a crucial role in the accuracy and stability of the resolution model. The test scenario of this embodiment is a complex urban environment, such as areas obstructed by tall buildings and under overpasses, as well as open environments. Within this range, in addition to multiple sets of dynamic RTK positioning experimental data, this embodiment also acquired GPS / BDS / Galileo multi-system observation data (including L1 / L5 / B1I / B2a / E1 / E5a ​​frequency points) collected by a low-cost multi-frequency GNSS receiver (such as U-blox, MXT906AM), as shown in Table 1 (Table 1 is the parameter table of low-cost GNSS receiver observation data). Using single-system (GPS) and multi-system (GPS / BDS / Galileo) observation data, this embodiment can construct ambiguity resolution sample sets in different scenarios to train and test the BIE-based integer ambiguity resolution method.

[0047]

[0048] This embodiment also conducted comparative tests on the Lappland distribution-based improved BIE method (LBIE) and typical integer ambiguity resolution methods such as traditional integer least squares, standard Gaussian distribution BIE (GBIE), and partial ambiguity fixation (PAR-ILS) to verify whether LBIE has excellent fixation rate and positioning accuracy. In the tests, low-cost GNSS observation data collected in an open environment (U-blox receiver L1 / L5 frequency points, MXT906AM receiver B1I / B2a frequency point data) were used as samples to train the model parameters (Lappland distribution scaling factor). The ADOP threshold was set to 0.0055. The test model was evaluated for integer ambiguity fixation rate and positioning accuracy (RMS values ​​in the E / N / U directions) in complex urban environments (high-rise buildings obstructing view, under overpasses). If a method can maintain a high fixation rate and low positioning error in complex environments without parameter readjustment, it indicates that the method has good environmental adaptability and can reduce the debugging workload of low-cost RTK multi-scenario applications.

[0049] Please see Figure 3 This is a schematic diagram comparing the positioning error sequences of different algorithms in complex environments according to embodiments of the present invention. Figure 3 The average values ​​of test indicators under complex urban environments (high-rise building obstruction + under elevated bridges) were statistically analyzed, and the data in Tables 2 and 3 below were obtained.

[0050]

[0051] Table 2 shows that the LBIE method exhibits superior performance across almost all performance metrics. The method achieves an ambiguity fixation rate of 96.25%, with RMS values ​​of 0.042m in the E direction, 0.038m in the N direction, and 0.051m in the U direction. These values ​​are the best among all compared solution methods, representing a 14.15 percentage point improvement in fixation rate compared to ILS's 82.10%. The RMS values ​​in the E, N, and U directions are reduced by 35.38%, 34.48%, and 29.17% respectively compared to GBIE's 0.065m, 0.058m, and 0.072m. This demonstrates that the LBIE method not only achieves high-precision ambiguity fixation in low-cost RTK but also maintains stable solution performance in complex environments, exhibiting excellent environmental adaptability.

[0052] Further testing of the specific performance of the method in various complex scene sub-regions shows that LBIE has excellent average performance in multiple sub-regions, especially in terms of ambiguity fixation rate, E-direction RMS value, N-direction RMS value and U-direction RMS value, where it consistently maintains an advantage over other comparative methods.

[0053]

[0054] Table 3 compares the localization accuracy of different algorithms in two typical complex scenarios: one is a tree-shaded scenario, characterized by intermittent occlusion and weak multipath, and the other is a Yangtze River Bridge scenario, characterized by strong geometric changes, significant multipath, and occlusion. Table 3 provides the accuracy metrics for PAR-ILS and PAR-BIE in the two scenarios, calculating the root mean square (RMS) or equivalent error values ​​in the E / N / U directions, and listing the overall 3D accuracy.

[0055] Overall comparison shows that in the shaded tree scene, PAR-BIE has smaller accuracy in the E / NE / N planes than PAR-ILS, and the U-direction error also decreases; in the Yangtze River Bridge scene, as occlusion and multipath intensify, the accuracy of both algorithms decreases, but PAR-BIE's plane and elevation errors increase less, and it maintains its advantage in overall 3D accuracy.

[0056] Table 3 is used for supplementary verification: Under different occlusion and geometric configurations, PAR-BIE based on Laplacian weight fusion and partial ambiguity fixing strategy is more robust to anomalous observations and heavy-tailed residuals, and can maintain higher positioning accuracy in complex environments.

[0057] Please see Figure 4This is a schematic diagram comparing the convergence time of the algorithm in different scenarios according to the embodiments of the present invention. It shows the convergence time comparison of two parallel fixed strategies in different scenarios under the two antenna configurations of Dingyao antenna and ceramic patch antenna. The left bar represents the parallel integer least squares strategy (PAR-ILS), and the right bar represents the parallel optimal integer equivariant estimation strategy (PAR-BIE).

[0058] For Dingyao antennas: In scenarios one and two, the convergence time of PAR-BIE is lower than or close to that of PAR-ILS; in scenario three, PAR-BIE further shortens the convergence time compared to PAR-ILS; in scenario four, where the obstruction is more severe or the multipath is more obvious, the convergence time of both increases, but the difference in magnitude is still distinguishable.

[0059] For ceramic patch antennas: In scenarios one through three, PAR-BIE exhibits a shorter convergence time compared to PAR-ILS; in scenario four, the convergence times of the two methods are similar but remain comparable, reflecting the stability assessment under low-cost antenna conditions.

[0060] according to Figure 4 It can be seen that: under different electromagnetic environments and geometric configurations, i.e. scenarios one to four, and different hardware levels, this embodiment specifically uses a high-performance external antenna and a low-cost patch antenna, and adopts a parallel ambiguity fixing process for comparative testing, recording the epoch accumulation time required to first achieve a fixed ambiguity and output a stable positioning solution. Figure 4 The two sets of bar charts correspond to the same data input, the same multi-system fusion and quality monitoring conditions, with only the fixed solution strategy being different, in order to reflect the difference in convergence speed between the parallel BIE strategy and the parallel ILS strategy.

[0061] This invention also provides a real-time dynamic positioning integer ambiguity resolution system based on BIE, the system comprising: The GNSS observation data filtering module is used to filter GNSS observation data. It removes abnormal satellites from the GNSS observation data based on satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset. The fuzziness decorrelation processing module is used to perform decorrelation processing on the initially screened fuzziness subset to obtain a decorrelated fuzziness subset. A multi-level fuzziness filtering and optimization module is used to dynamically optimize the decorrelated fuzziness subset using a multi-level filtering mechanism to obtain a fuzziness candidate group; The fusion weight calculation and ambiguity fixing module is used to calculate the weight of each candidate solution in the ambiguity candidate group based on the weight allocation model improved by Laplace distribution, perform probability weighted fusion on the ambiguity candidate group to obtain fused ambiguity, fix the fused ambiguity using the ILS method, and when the ILS fixing fails, the BIE method is used to perform weighted averaging on the fused ambiguity to obtain fixed ambiguity. The positioning result calculation and multi-system fusion correction module is used to calculate the initial positioning result based on the fixed ambiguity, optimize the initial positioning result through a dynamic satellite filtering strategy to obtain an optimized positioning result, and combine multi-system data fusion to correct the error influence of the optimized positioning result to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

[0062] Specifically, this embodiment presents a real-time dynamic positioning integer ambiguity resolution system based on BIE. It constructs five functional modules: GNSS observation data filtering, ambiguity decorrelation processing, multi-level filtering optimization, fusion weight calculation and ambiguity fixing, and positioning result calculation and multi-system fusion correction. Through the coordinated cooperation of these modules, it achieves full-process automation of data preprocessing, quality control, candidate group optimization, and weight fusion. It improves the switching mechanism between weight allocation and the ILS-BIE hybrid fixing strategy by incorporating Laplace distribution. Furthermore, through closed-loop feedback control of multi-system data fusion and jump detection, it solves the problems of unstable ambiguity fixing and large fluctuations in positioning accuracy in traditional RTK systems under complex environments, thereby improving the overall reliability and environmental adaptability of the real-time dynamic positioning system.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time dynamic positioning integer ambiguity resolution based on BIE, characterized in that, Includes the following steps: The GNSS observation data is filtered, and abnormal satellites in the GNSS observation data are removed according to the satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset; The pre-selected fuzzy subset is subjected to decorrelation processing to obtain a decorrelated fuzzy subset; A multi-level filtering mechanism is used to dynamically optimize the decorrelated fuzziness subset to obtain a candidate group of fuzziness. Based on the weight allocation model improved by Laplace distribution, the weights of each candidate solution in the candidate group of fuzziness screening are calculated, and the candidate group of fuzziness screening is probabilistically weighted and fused to obtain the fused fuzziness. The fused fuzziness is fixed by the ILS method. When the ILS fixation fails, the fused fuzziness is weighted and averaged by the BIE method to obtain the fixed fuzziness. The initial positioning result is calculated based on the fixed ambiguity. The initial positioning result is optimized by a dynamic satellite filtering strategy to obtain an optimized positioning result. The error influence of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

2. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 1, characterized in that, The multi-level screening mechanism includes a coarse screening stage, a fine screening stage, and a candidate group generation stage, wherein... The coarse screening stage includes: removing satellites with a PDOP value higher than 3.0 based on the satellite geometry configuration PDOP value, and removing satellites with a carrier-to-noise ratio lower than 35dBHz based on the signal-to-noise ratio, to obtain a coarsely screened subset of decorrelated ambiguity. The fine screening stage includes: evaluating the bootstrapping success rate of the decorrelated ambiguity subset after coarse screening, calculating the fixed probability of each ambiguity, screening ambiguities with a fixed probability higher than 90%, and simultaneously deleting low-quality ambiguities in descending order of Z-transform decorrelated variance, dynamically adjusting the size of the ambiguity subset to 6-10, and obtaining the finely screened ambiguity subset. The candidate group generation stage includes: setting the maximum number of candidate groups to 100 based on the finely filtered fuzzy subset, calculating the weight of each candidate solution using a weight allocation model improved by Laplace distribution, and obtaining the filtered fuzzy candidate groups.

3. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 2, characterized in that, The Bootstrapping success rate assessment in the fine screening stage is based on the conditional variance calculation of the fixed probability of ambiguity, and high fixed probability ambiguities are screened by evaluating the strength of the GNSS mathematical model. The Z-transform decorrelation variance screening sorts the ambiguity variance-covariance matrix by decorrelation sorting, prioritizing the retention of high-quality ambiguities with small variances and deleting low-quality ambiguities with large variances. When the size of the ambiguity subset exceeds 10, the ambiguities are deleted from largest to smallest according to the decreasing correlation variance and the ADOP value is recalculated to ensure that the ADOP value is less than 0.0055. When the size of the ambiguity subset is less than 6, GNSS observation data is re-acquired and the ambiguities corresponding to the satellites that meet the conditions are supplemented and filtered. During the candidate group generation phase, the Laplace distribution improved weight allocation model replaces the exponentially decaying weights of the traditional Gaussian distribution with linearly decaying weights.

4. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 1, characterized in that, The probability-weighted fusion includes an ILS fixed judgment phase and a BIE weighted average phase; The ILS fixing judgment stage includes: fixing the ambiguity of the selected ambiguity candidate group using the ILS method, and determining whether the ILS fixing is successful by combining the Ratio test and the F test. When the Ratio test threshold is greater than 3.0 and the significance level of the F test is greater than 0.05, the ILS fixing result is output to obtain the fixed ambiguity. The BIE weighted averaging stage includes: when ILS fixation fails, calculating the likelihood function weight of the fused ambiguity based on the Laplace distribution probability density function, replacing the traditional Gaussian distribution exponential decay model with a linear decay weight model, and performing a weighted average on all candidate solutions of the fused ambiguity to obtain the fixed ambiguity after BIE weighted averaging. Fixing the fused ambiguities includes: combining the PAR-BIE strategy, partial ambiguity fixing, and optimal integer equivariant estimation; fixing satellite combinations with PDOP values ​​less than 2.0 and Bootstrapping fixing probability greater than 95%; when the size of the ambiguity subset exceeds 10, deleting ambiguities from largest to smallest according to the Z-transform to reduce correlation variance, and recalculating the ADOP value; when the size of the ambiguity subset is less than 6, re-collecting GNSS observation data and supplementing the ambiguities corresponding to the satellites.

5. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 4, characterized in that, In the BIE weighted average stage, the scaling factor of the probability density function of the Laplace distribution is set to 4. Based on the scaling factor, the geometric distance between each candidate solution and the floating-point solution is calculated, and the geometric distance is converted into a weight value through the Laplace distribution function. The linear decay weight model adopts the double exponential decay characteristic of the Laplace distribution, assigning relatively higher weights to candidate solutions that are far from the floating-point solution and relatively lower weights to candidate solutions that are close to the floating-point solution. By calculating the PDOP value and Bootstrapping fixation probability of each satellite combination, satellite combinations with a PDOP value less than 2.0 and a Bootstrapping fixation probability greater than 95% are selected for priority fixation. When the size of the ambiguity subset exceeds 10, ambiguity parameters are deleted one by one in descending order of the Z-transform downcorrelation variance value in the ambiguity variance-covariance matrix. After deletion, the ADOP value is recalculated and compared with the second preset threshold. When the ADOP value is less than the second preset threshold, the current ambiguity subset is retained. When the size of the ambiguity subset is less than 6, GNSS observation data is reacquired and satellites are added.

6. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 1, characterized in that, The initial positioning result is calculated based on the fixed ambiguity. The initial positioning result is then optimized using a dynamic satellite filtering strategy to obtain an optimized positioning result. Finally, the error impact of the optimized positioning result is corrected by multi-system data fusion to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is repeated, including: Position is calculated using the carrier phase observation equation. The PDOP threshold is set to 2.0 and the carrier-to-noise ratio threshold is set to 35 dBHz. Satellites exceeding the threshold range are removed to obtain a filtered satellite set. The optimized positioning result is obtained by recalculating based on the filtered satellite set. The multi-system data fusion is achieved by fusing observation data from GPS, BDS, and Galileo systems using a weighted least squares method. The weighting coefficients of the GPS, BDS, and Galileo system observation data are set according to the observation noise variance of each system to correct the optimized positioning result and obtain the final positioning result. The first preset threshold is set to 0.05m. When the absolute value of the coordinate difference between the final positioning result and the previous epoch positioning result exceeds the first preset threshold, the calculation is performed again.

7. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 6, characterized in that, The carrier phase observation equation is based on a double-difference observation model to eliminate receiver clock bias and satellite clock bias. The filtered satellite set is determined by satellite continuity between epochs by recording the PDOP value and carrier-to-noise ratio data of satellites in two consecutive epochs. When a satellite continuously meets the conditions of PDOP value less than 2.0 and carrier-to-noise ratio greater than 35dBHz, the satellite is included in the filtered satellite set. The weighted least squares method includes the following weighting coefficients for the GPS system: the pseudorange noise variance system weighting coefficient is 0.3, the carrier phase noise variance system weighting coefficient is 0.001, and the weighting coefficients for the BDS and Galileo systems are adjusted proportionally according to the signal quality. The calculation of the absolute value of the coordinate difference includes the coordinate component differences in the three directions of E, N, and U; The recalculation process includes re-filtering GNSS observation data, re-performing decorrelation processing, re-filtering at multiple levels, and re-performing BIE probability weighted fusion until the final positioning result meets the accuracy requirements.

8. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 1, characterized in that, The process of filtering GNSS observation data involves removing abnormal satellites based on satellite elevation angle, carrier-to-noise ratio, and observation residuals to obtain a preliminary ambiguity subset, including: Satellites with an elevation angle below 15° and a carrier-to-noise ratio below 35dBHz were removed to obtain the observation data after filtering by elevation angle and carrier-to-noise ratio; The observation data after being filtered by elevation angle and carrier-to-noise ratio are subjected to robust adaptive Kalman filtering, the observation residuals of each satellite are calculated, and satellites with observation residuals exceeding twice the root mean square error are removed to obtain the preliminary ambiguity subset.

9. The method for real-time dynamic positioning integer ambiguity resolution based on BIE as described in claim 1, characterized in that, The step of performing decorrelation processing on the initially screened fuzzy subset to obtain a decorrelated fuzzy subset includes: The initially screened ambiguity subset is subjected to Z-transform decorrelation processing, and the ambiguity variance-covariance matrix is ​​calculated using the LAMBDA method to obtain the ambiguity variance-covariance matrix after decorrelation processing. The ADOP value is calculated based on the ambiguity variance-covariance matrix after the downcorrelation processing. The ADOP value is compared with a second preset threshold. When the ADOP value is lower than the second preset threshold, the downcorrelation ambiguity subset is obtained. When the ADOP value is higher than the second preset threshold, the ambiguity subset is re-optimized.

10. A BIE-based real-time dynamic positioning integer ambiguity resolution system, used to execute the BIE-based real-time dynamic positioning integer ambiguity resolution method as described in any one of claims 1-9, characterized in that, The system includes: The GNSS observation data filtering module is used to filter GNSS observation data. It removes abnormal satellites from the GNSS observation data based on satellite elevation angle, carrier-to-noise ratio and observation residuals to obtain a preliminary ambiguity subset. The fuzziness decorrelation processing module is used to perform decorrelation processing on the initially screened fuzziness subset to obtain a decorrelated fuzziness subset. A multi-level fuzziness filtering and optimization module is used to dynamically optimize the decorrelated fuzziness subset using a multi-level filtering mechanism to obtain a fuzziness candidate group; The fusion weight calculation and ambiguity fixing module is used to calculate the weight of each candidate solution in the ambiguity candidate group based on the weight allocation model improved by Laplace distribution, perform probability weighted fusion on the ambiguity candidate group to obtain fused ambiguity, fix the fused ambiguity using the ILS method, and when the ILS fixing fails, the BIE method is used to perform weighted averaging on the fused ambiguity to obtain fixed ambiguity. The positioning result calculation and multi-system fusion correction module is used to calculate the initial positioning result based on the fixed ambiguity, optimize the initial positioning result through a dynamic satellite filtering strategy to obtain an optimized positioning result, and combine multi-system data fusion to correct the error influence of the optimized positioning result to obtain the final positioning result. When the jump between the final positioning result and the positioning result of the previous epoch exceeds a first preset threshold, the calculation is performed again.

Images