A method and system for evaluating data quality of Beidou

Abstract

The application discloses a Beidou data quality evaluation method and system, and relates to the field of high-precision positioning technology.The method comprises the following steps: acquiring a satellite post-check residual sequence; performing asymmetric quantization on the sequence, calculating a third central moment through a sliding time domain window, and obtaining a single-satellite noise skewness coefficient value representing the skewness degree of statistical distribution; then, based on the skewness coefficients of all satellites and the sky geometric positions thereof, spatialized weighted fusion is performed to extract a two-dimensional skewness torque pseudo-observation which can macroscopically represent the asymmetry of the electromagnetic environment around the receiver.The application also provides an application verification scheme, which takes the skewness torque as a pseudo-observation, performs online estimation and closed-loop compensation on systematic positioning bias in the augmented Kalman filtering framework, and finally outputs corrected high-precision positioning results.The application breaks through the limitation of traditional methods which can only evaluate the noise size, and can effectively improve the accuracy and reliability of Beidou positioning in complex environments.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of high-precision positioning technology, and particularly relates to a Beidou data quality evaluation method and system. BACKGROUND

[0002] In the application fields of high-level automatic driving and high-precision mapping, there are extremely strict performance requirements for positioning, navigation and timing services based on global navigation satellite systems such as Beidou. However, in a complex environment such as a city canyon with high-rise buildings, the propagation of global navigation satellite system signals faces severe challenges. On the one hand, satellite signals are easily blocked by buildings, resulting in a decrease in the number of visible satellites; on the other hand, the multipath effect caused by the reflection or diffuse reflection of signals on building surfaces will cause signals carrying error time delay to superimpose on direct signals, causing complex dynamic errors in pseudo-range, carrier phase and other raw observations.

[0003] For example, when the west side of an automatic driving vehicle is adjacent to a continuous high-rise glass curtain wall, and the east side is an open park, a large number of multipath signals with strong energy will always come from the west sky. If these multipath signals always make the pseudo-range measurement value systematically larger, the long-term statistical distribution of the pseudo-range observation noise will show a significant positive skew. This noise with the characteristics of a statistically non-zero mean and significant skew will continuously “pull” the positioning solution result to an incorrect direction, thereby causing a persistent systematic positioning deviation of the Kalman filter that is difficult to eliminate through the traditional filtering framework. According to the public research and field measurement data in the related field, in a typical city canyon environment, the systematic positioning deviation caused by the asymmetric multipath effect can be as high as 5 to 15 meters, and in extreme cases, it can be even larger. An error of this magnitude is completely unacceptable for automatic driving lane keeping functions that require sub-meter or even centimeter-level lateral positioning accuracy, and can directly lead to the vehicle deviating from the lane and colliding with obstacles, and other catastrophic consequences.

[0004] To deal with the multipath effect, the existing technology mainly suppresses it from the following aspects:

[0005] Hardware level: Use a choke coil antenna or an antenna array with multipath suppression capability to physically weaken the reception of multipath signals. However, such hardware is high in cost and large in size, and is difficult to be widely applied to cost- and space-sensitive vehicle-mounted devices.

[0006] Signal processing level: In the receiver baseband processing, advanced delay-locked loop (DLL) technologies such as narrow correlation can be used to suppress part of the multipath signals, but for multipath signals with longer delays or similar delays to direct signals, the suppression effect will decrease sharply.

[0007] Navigation filter level: In state estimation algorithms such as Kalman filter, it is a common practice to dynamically adjust the weight of observation noise according to the carrier-to-noise ratio of satellite signals. When the carrier-to-noise ratio of a satellite signal decreases, the system will increase the corresponding observation noise variance, thereby reducing the contribution of the satellite data in positioning solution. This method to some extent acknowledges the decline in data quality.

[0008] However, the above prior art is based on a common but often not true assumption in challenging environments: the statistical distribution of observation noise caused by factors such as multipath is symmetric (usually assumed to be a zero-mean Gaussian distribution). SUMMARY

[0009] The present application provides a Beidou data quality evaluation method and system to solve the problem that the prior art is limited to analyzing the second-order statistical moments (variance) of observation noise when evaluating data quality. This means that traditional methods (such as Kalman filter based on carrier-to-noise ratio weighting) can only judge the "size" or "dispersion" of noise, but lack effective perception and quantification means for the third-order statistical moments (skewness) that can reflect the systematic bias of errors.

[0010] In view of the above problems, in a first aspect, the present application provides a Beidou data quality evaluation method, comprising the following steps:

[0011] Obtaining the original observation data of at least one Beidou satellite for a navigation receiver within a plurality of consecutive positioning epochs;

[0012] Using a main positioning filter to process the original observation data to calculate the posteriori residuals of the at least one Beidou satellite within the plurality of consecutive positioning epochs, forming a posteriori residual sequence;

[0013] Performing asymmetric quantization processing on the posteriori residual sequence to obtain a single-satellite noise skewness coefficient value representing the skewness degree of the statistical distribution of the posteriori residual sequence;

[0014] Based on the single-satellite noise skewness coefficient value of the at least one Beidou satellite and its geometric position in the sky, performing spatialized fusion processing to extract a skewness moment pseudo-observation quantity that can macroscopically represent the asymmetry of the electromagnetic environment around the navigation receiver.

[0015] In a second aspect, the present application also provides a Beidou data quality evaluation system, comprising:

[0016] a data acquisition and processing unit configured to acquire raw observation data of at least one Beidou satellite by a navigation receiver at a plurality of consecutive positioning epochs, and to process the raw observation data by using a main positioning filter to calculate a posteriori residuals of the at least one Beidou satellite at the plurality of consecutive positioning epochs, thereby forming a sequence of a posteriori residuals;

[0017] a data quality assessment unit connected with the data acquisition and processing unit, and configured to perform asymmetric quantization on the sequence of a posteriori residuals to obtain a single-satellite noise skewness coefficient value representing a skewness degree of statistical distribution of the sequence of a posteriori residuals, and to perform spatialized fusion processing based on the single-satellite noise skewness coefficient value of the at least one Beidou satellite and a geometric position of the at least one Beidou satellite in the sky, thereby extracting a skewness moment pseudo-observation quantity capable of macroscopically representing the asymmetry of the electromagnetic environment around the navigation receiver;

[0018] an application verification unit connected with the data quality assessment unit and the data acquisition and processing unit, wherein the data quality assessment unit is further configured to output the skewness moment pseudo-observation quantity to the application verification unit, and the application verification unit is configured to receive the skewness moment pseudo-observation quantity, estimate and compensate a systematic positioning bias caused by the asymmetry of the electromagnetic environment in an augmented state filtering framework, and output a final positioning result, and the application verification unit is further configured to feed back an estimation result of the systematic positioning bias to the main positioning filter in the data acquisition and processing unit, so as to realize closed-loop correction.

[0019] The technical scheme provided in the application has at least the following technical effects or advantages:

[0020] The application introduces the "asymmetry" (skewness, third moment) of observation noise as a brand-new data quality assessment dimension. This enables the system to perceive and quantify the "directivity" of observation noise, and fundamentally improves the insight into the characteristics of degraded data in a complex environment.

[0021] The application realizes macroscopic and quantitative perception of the direction and intensity of the asymmetry of the overall electromagnetic environment around the receiver by constructing a "sky skewness map" and extracting a "skewness moment pseudo-observation quantity", and spatially and structurally fusing the skewness information of all discrete satellites. The skewness moment is an environmental characteristic index with clear physical meaning.

[0022] The application can accurately estimate and compensate the systematic positioning deviation caused by the asymmetric multipath effect in real time by taking the skewness torque as a pseudo-observation in an augmented filtering framework. This directly solves the persistent positioning deviation problem that the traditional method cannot handle, and can significantly reduce the systematic error of several to tens of meters caused by asymmetric multipath in challenging environments such as urban canyons, thereby greatly improving the accuracy, reliability and continuity of high-precision positioning results.

[0023] When the skewness torque is compensated, the application introduces an adaptive confidence model coupled with the satellite geometry dilution of precision (GDOP) and the number of visible satellites. This model can adjust the degree of trust of the system for skewness torque pseudo-observations online, ensuring the robustness and stability of the evaluation and compensation algorithm in poor satellite geometry distribution or fewer visible satellites, and enhancing the overall performance and environmental adaptability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A flowchart of a Beidou data quality evaluation method according to the application;

[0025] Figure 2 A system architecture diagram for evaluating the quality of Beidou data. DETAILED DESCRIPTION

[0026] The application relates to a Beidou data quality evaluation method and system to solve the problem that the prior art method for evaluating data quality is limited to analyzing the second-order statistical moment (variance) of observation noise. This means that the traditional method (such as Kalman filtering based on carrier-to-noise ratio weighting) can only determine the "size" or "dispersion degree" of the noise, and completely lacks effective sensing and quantization means for the third-order statistical moment (skewness) that can reflect the systematic deviation of the error.

[0027] The above technical solutions will be described in detail below in combination with the accompanying drawings and specific embodiments, so that the above technical solutions can be better understood. Obviously, the described embodiments are only part of the embodiments of the application, not all embodiments of the application, and it should be understood that the application is not limited to the example embodiments for explaining the application. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application. In addition, it should be noted that, for convenience of description, only parts related to the application are shown in the drawings, not all.

[0028] Please refer to Figure 1 A Beidou data quality evaluation method, comprising the following steps:

[0029] Obtaining raw observation data of a navigation receiver for at least one Beidou satellite in a plurality of continuous positioning epochs;

[0030] Processing the raw observation data by using a main positioning filter to calculate post-processed residuals of the at least one Beidou satellite in the plurality of continuous positioning epochs, to form a post-processed residual sequence;

[0031] Performing asymmetric quantization processing on the post-processed residual sequence to obtain a single-satellite noise skewness coefficient value for representing a skewness degree of statistical distribution of the post-processed residual sequence;

[0032] Performing spatialized fusion processing based on the single-satellite noise skewness coefficient value of the at least one Beidou satellite and a geometric position of the at least one Beidou satellite in the sky to extract a skewness torque pseudo-observation quantity capable of macroscopically representing an asymmetry of an electromagnetic environment around the navigation receiver.

[0033] The implementation of the Beidou data quality evaluation method covers the complete engineering life cycle from offline configuration and calibration before system deployment to real-time online processing in operation and data utilization and traceability after operation, and exhibits high engineering practicability.

[0034] Stage one: system deployment and offline parameter calibration

[0035] Before deploying the method described in the present application to actual receiver firmware or vehicle-mounted high-precision positioning units, one-time system software integration and offline calibration of key engineering parameters are required.

[0036] Software module integration and interface definition: the method described in the present application is a pure software algorithm module in physical form. The module will be compiled and integrated into receiver firmware or upper-layer application processing software. A clear bidirectional data interface needs to be established with the original main positioning filter module (for example, an implementation based on an extended Kalman filter) in the system. The interface definition should at least include: an interface for one-way transmission of post-processed residual sequences from the main positioning filter to the module of the present application; and an interface for one-way feedback of positioning bias state estimation from the module of the present application to the main positioning filter.

[0037] Offline calibration of key engineering parameters: the present application contains several key engineering parameters whose values directly affect system performance. The determination of these parameters is not dependent on repeated trial and error or empirical guess in the field, but is completed through a standard and repeatable offline calibration experiment, which constitutes an important cornerstone of the reproducibility of the technical solution of the present application.

[0038] Stage two: real-time online processing cycle

[0039] After the system is deployed and initialized, at each positioning epoch (e.g. usually once or ten times per second in vehicle applications), the system will automatically, cyclically execute a complete real-time online processing flow. This flow is the core operational logic of the invention. The macroscopic data flow and processing flow of this cycle are as follows:

[0040] At each positioning epoch, the system automatically and cyclically executes a complete real-time online processing flow. This flow begins with the reception of the epoch raw observation data packet output by the receiver RF front-end and baseband processing unit. First, the main positioning filter performs standard state prediction and update on this data packet, while calculating the receiver state and generating the key single-satellite post-update residual value for each satellite. Then, the data quality assessment unit described in the invention receives these post-update residual values and performs its core quality assessment task: it calculates the single-satellite noise skewness coefficient in real-time through sliding window statistics, then constructs a dynamic sky skewness map, and finally extracts the skewness torque pseudo-observation that macroscopically characterizes the environmental asymmetry from it. Then, the application verification unit uses this skewness torque pseudo-observation as an observation information to update and optimally estimate the systematic positioning bias state vector in an augmented Kalman filter framework. Finally, in the closed-loop compensation stage, the main positioning filter receives the feedback of the positioning bias state vector, directly corrects the receiver's position state in its operational flow, and finally outputs a compensated, higher-precision final positioning result.

[0041] Stage three: Subsequent data utilization, traceability and model evolution

[0042] To ensure the maintainability, interpretability and continuous optimization capability of the system, the design of the invention also includes a recording and utilization mechanism for key intermediate data.

[0043] During system operation, the following key data of each positioning epoch, along with high-precision timestamps, will be recorded in a structured and efficient format (e.g. ProtocolBuffers or custom binary format) to non-volatile memory (such as vehicle SSD): raw observation data summary, standard positioning result, noise skewness coefficient value of each satellite output by the method described in the invention, calculated two-dimensional skewness torque pseudo-observation vector, estimated two-dimensional positioning bias state vector, and finally compensated positioning result.

[0044] When positioning performance anomalies occur (e.g. vehicle reports positioning accuracy exceeding limits) or offline evaluation of navigation performance on a specific section is needed, operation and maintenance or research and development engineers can extract the log data of this period. By plotting and analyzing the time series of "skewness torque" and "positioning bias", the system's perception of the asymmetry of the surrounding electromagnetic environment at that time can be accurately reproduced. This provides an unprecedented diagnostic capability:

[0045] Scenario 1: If the logs show that the “skew torque” consistently points east and the “position bias” also consistently converges to a value pointing east, it can be highly confident that there is one or more strong reflectors on the west side of the road segment. This provides valuable data input for the production and verification of high-precision maps, especially “feature maps” or “risk maps”.

[0046] Scenario 2: If the logs show that the positioning results have large jumps, but at the same time the “skew torque” and “position bias” are both around zero, it can be initially judged that the positioning problem is not caused by environmental asymmetry, but more likely caused by other reasons such as receiver's own transient failure, unsuccessful cycle slip repair, or individual satellite's ephemeris / clock error anomaly, thereby greatly narrowing the scope of troubleshooting.

[0047] The long-term accumulation of log data covering various road and weather conditions can serve as a large-scale and richly annotated “real-world data set”. This data set can be used to re-evaluate and optimize the system parameters of offline calibration (such as “mapping scale factor”), and even be used to train more complex, machine learning-based nonlinear mapping models to replace the current preferred linear model, thereby realizing the continuous iteration and performance evolution of the algorithm model.

[0048] Further, the implementation of the core technology of the present application.

[0049] Step 1: Acquisition and dynamic management of multi-epoch post-processing residual sequence

[0050] Data form A, epoch original observation data packet, defined as the original measurement data set related to multiple satellites obtained from the receiver baseband signal processing unit at any discrete positioning epoch without any upper layer filtering processing.

[0051] Data structure: a collection containing multiple sub-structures, where each sub-structure uniquely corresponds to a currently visible and tracked Beidou satellite. The sub-structure is named single-satellite original observation structure, which at least contains the following fields:

[0052] Satellite unique identifier (integer, e.g. satellite PRN number);

[0053] C1 pseudo-range observation value (double-precision floating point type, unit: meter);

[0054] L1 carrier phase observation value (double-precision floating point type, unit: cycle);

[0055] D1 Doppler shift (single-precision floating point type, unit: hertz);

[0056] S1 carrier-to-noise ratio (single-precision floating point, unit: decibel hertz);

[0057] (optionally, similar observations of other frequency points, such as B2, B3 frequency points);

[0058] Generation process of post-filter residual (taking an extended Kalman filter as an example).

[0059] The main positioning filter (in this embodiment, an extended Kalman filter) first performs its standard state prediction step. By using the post-state estimation vector of the last positioning epoch and the system dynamic model (described by a state transition matrix), the prior state estimation vector of the current positioning epoch is predicted through matrix operation.

[0060] Subsequently, the main positioning filter performs its standard state update step. By using the epoch original observation data packet of the current positioning epoch and the observation model (described by an observation Jacobian matrix), the post-state estimation vector of the current positioning epoch is calculated by calculating the Kalman gain and combining the prior state estimation vector. This post-state is the optimal estimation of the receiver state (position, velocity, clock bias, etc.) after fusing all current observation information.

[0061] After obtaining the post-state estimation vector, the system re-calculates a theoretically observed measurement value for each satellite participating in the solution by using the optimal post-state estimation vector. This value is called the post-theoretical observation value. Then, the actual observation value (for example, the C1 pseudo-range observation value) of the satellite is subtracted from the re-calculated post-theoretical observation value.

[0062] Data form B, single-satellite post-filter residual value, which is defined as a scalar value calculated at the current positioning epoch for a specific satellite and a specific observation type (preferably pseudo-range observation in this embodiment) that can best reflect unmodeled errors (especially multipath effects).

[0063] Data type: double-precision floating point.

[0064] Step 2: Online quantification of single-satellite noise skewness coefficient based on sliding time domain window For each satellite, an index that quantitatively describes the degree of skewness of the statistical distribution of the post-filter residual sequence, the single-satellite noise skewness coefficient value, is calculated in real time and independently.

[0065] Single-satellite residual time domain window: for each currently visible and tracked Beidou satellite, an independent data queue structure with a first-in-first-out feature is instantiated and maintained in the dynamic memory of the system.

[0066] Data structure implementation: In a preferred embodiment, the data structure is implemented using a circular array (CircularArray) to achieve the highest access efficiency. The capacity of the circular array is set to a preset, tunable integer, called the preset queue length. In the data structure body, in addition to the array storing the residual values, it also needs to include an index pointing to the current head element, an index pointing to the current tail element, and a counter recording the number of valid elements in the current queue.

[0067] At each positioning epoch, for each visible satellite, the system will perform the following series of micro-logic operations:

[0068] Push the newly calculated single-satellite a posteriori residual value of this positioning epoch into the tail of the single-satellite residual time-domain window data structure corresponding to it. This operation is specifically to write the new residual value into the array position pointed to by the tail index, and then move the tail index forward by one bit (if it reaches the end of the array, it wraps around to the beginning of the array).

[0069] After the enqueue operation, determine whether the valid element number counter of the single-satellite residual time-domain window has exceeded the preset queue length. If so, move the head index forward by one bit (again, if it reaches the end of the array, it wraps around), which is logically equivalent to discarding the oldest residual value in the queue.

[0070] Call a software function module named sliding window skewness calculation unit and pass the memory address of the current single-satellite residual time-domain window as a parameter. This unit is responsible for calculating the sample skewness coefficient for all valid residual values stored in the current window.

[0071] Sample skewness coefficient calculation method: the calculation logic inside the sliding window skewness calculation unit includes the following steps:

[0072] First step: calculation of first-order raw moment (sample mean). Initialize a local variable of double-precision floating-point type named first-order moment accumulator, with an initial value of zero. Start a loop to iterate through all valid residual values in the current single-satellite residual time-domain window. In each iteration of the loop, add the current residual value to the first-order moment accumulator. After the loop ends, divide the final value of the first-order moment accumulator by the number of valid elements in the current window to obtain the sample mean.

[0073] Second step: calculation of second-order and third-order central moments. Initialize two local variables of double-precision floating-point type named second-order central moment accumulator and third-order central moment accumulator, with initial values of zero. Start a loop again to iterate through all valid residual values in the window. In each iteration of the loop:

[0074] First, the difference between the current residual value and the sample mean calculated in the previous step is calculated. This difference is called the deviation.

[0075] Then, the square of the deviation is calculated and the result is accumulated into the second central moment accumulator.

[0076] Next, the cube of the deviation is calculated and the result is accumulated into the third central moment accumulator.

[0077] Third step: final determination of unbiased standard deviation and third central moment. After the end of the second loop:

[0078] The final value of the second central moment accumulator is divided by the number of valid elements minus one, and the result is the statistically recognized unbiased sample variance.

[0079] The square root of the unbiased sample variance is taken to obtain the unbiased sample standard deviation.

[0080] The final value of the third central moment accumulator is divided by the number of valid elements to obtain the third central moment.

[0081] Fourth step: final calculation of the normalized skewness coefficient. In order to eliminate the influence of sample size and the dimension of data itself on the estimation of skewness, and to obtain a universal and standardized evaluation index, the present application preferably uses a sample size corrected normalization transformation to obtain the final skewness coefficient value.

[0082] Sample size corrected normalized skewness coefficient: the operation relationship of the transformation is defined as follows: first, a sample size correction factor related to the number of valid elements in the current window is calculated. The calculation method of the sample size correction factor is as follows: take the square root of the product of the number of valid elements and "the number of valid elements minus one", and then divide the result by "the number of valid elements minus two". Then, divide the third central moment calculated in the previous step by the cube of the unbiased sample standard deviation. Finally, multiply the two calculation results to obtain the final value, which is the single star noise skewness coefficient value.

[0083] Data form C, single star noise skewness coefficient value, which is defined as a dimensionless, standardized, small sample corrected quantitative evaluation result of the asymmetry of the observation noise distribution of a specific satellite at the current positioning epoch.

[0084] Data type: double-precision floating point. Its value is positive, indicating that the noise distribution is skewed to the positive direction (usually corresponding to larger pseudorange); its value is negative, indicating that it is skewed to the negative direction; its value is close to zero, indicating that the distribution is close to symmetric.

[0085] Step three: Construction of the sky skewness map and extraction of the skewness moment pseudo-observation. All the satellite discrete and independent quality assessment information is spatialized and structured fused to extract the key vector information that can macroscopically represent the asymmetry of the overall electromagnetic environment around the receiver.

[0086] Definition of the data structure: Sky skewness map: a collection for storing the skewness assessment results and related geometric information of all currently visible and successfully processed satellites at the current positioning epoch.

[0087] Implementation of the data structure: In a preferred embodiment, the data structure is implemented using a hash table (HashMap) or dictionary (Dictionary) to support efficient satellite identifier-based lookup, insertion, and deletion operations. The hash table key (Key) is the satellite unique identifier (integer type). The hash table value (Value) is a content-rich structure body named single-satellite comprehensive state structure body, which at least contains the following fields:

[0088] Satellite azimuth (double-precision floating-point type, unit: degree, defined as the angle of clockwise rotation from the north direction)

[0089] Satellite elevation angle (double-precision floating-point type, unit: degree, defined as the angle with the local horizontal plane)

[0090] Single-satellite noise skewness coefficient value (double-precision floating-point type)

[0091] Extraction of the skewness moment pseudo-observation:

[0092] Step one: Initialization of local variables. At the beginning of the algorithm, four double-precision floating-point local variables are declared and their values are strictly initialized to zero. The four variables are:

[0093] East-west direction weighted skewness accumulator;

[0094] North-south direction weighted skewness accumulator;

[0095] Total weight accumulator;

[0096] Effective satellite counter.

[0097] Step two: Iterative calculation by traversing the sky skewness map. Start a loop to traverse each key-value pair in the current sky skewness map hash table. In each iteration of the loop, for the current satellite, perform the following series of sub-steps:

[0098] a. Data extraction: Obtain the single-satellite noise skewness coefficient value, satellite azimuth, and satellite elevation angle from the single-satellite comprehensive state structure body of the current satellite.

[0099] b. Contribution weight calculation: apply the unique Formula Two (Saturation-Weighted Model of Inverse Function of Elevation) to calculate the contribution weight of the current satellite.

[0100] Saturation-Weighted Model of Inverse Function of Elevation: the method of calculating the contribution weight is as follows: first, convert the satellite elevation angle from degrees to radians. Then, calculate the sine of the radian value. Next, add the sine value to a small positive real number called saturation stabilizer, which is used to prevent the denominator from being zero. Finally, take the reciprocal of the sum, and the result is the contribution weight of the current satellite.

[0101] c. Vectorized decomposition of the bias component: project and decompose the scalar form of the single-satellite noise bias coefficient value into a local East-North-Up (ENU) geographic coordinate system.

[0102] The calculation method of the east-west direction bias component is as follows: multiply the single-satellite noise bias coefficient value by the sine of the satellite azimuth angle.

[0103] The calculation method of the north-south direction bias component is as follows: multiply the single-satellite noise bias coefficient value by the cosine of the satellite azimuth angle.

[0104] d. Weighted accumulation operation: multiply the east-west direction bias component calculated in the previous step by the contribution weight, and accumulate the result to the east-west direction weighted bias accumulator. Multiply the north-south direction bias component by the contribution weight, and accumulate the result to the north-south direction weighted bias accumulator. Accumulate the contribution weight itself to the total weight accumulator. Increment the effective satellite counter by one.

[0105] Third step: validity judgment and final normalization calculation. After the traversal loop ends, first perform a validity judgment: check whether the value of the total weight accumulator is greater than a small positive threshold (for example, 1 times 10 to the power of -6), and whether the effective satellite counter is greater than or equal to a minimum satellite number threshold (for example, 4). This judgment aims to avoid producing meaningless or unstable calculation results in the case of too few visible satellites or all satellites being near the zenith, resulting in too small weights.

[0106] If the validity judgment passes, perform the final normalization calculation:

[0107] The final value of the east-west direction moment component is: divide the final value of the east-west direction weighted bias accumulator by the final value of the total weight accumulator.

[0108] The final value of the north-south direction moment component is: divide the final value of the north-south direction weighted bias accumulator by the final value of the total weight accumulator.

[0109] Data form: A two-dimensional vector, which is defined as the core quality assessment vector that can represent the direction and strength of the overall electromagnetic environment asymmetry around the receiver after synthesizing all the visible satellite noise asymmetry information at the current positioning epoch.

[0110] Data structure: An array or vector containing two elements: [Eastward moment component, North-South moment component].

[0111] Data type: Array of double-precision floating-point type.

[0112] Step four and step five: Construction of augmented filtering framework and closed-loop compensation based on quality assessment

[0113] Construction of augmented state Kalman filter:

[0114] The original n-dimensional state vector of the main positioning filter (in this embodiment, the extended Kalman filter) is extended. Specifically, two dimensions are added to the end of the state vector to form an (n+2)-dimensional augmented state vector. The two new state quantities are defined as:

[0115] Eastward systematic positioning bias;

[0116] North-South systematic positioning bias.

[0117] Augmentation of state transition and process noise model: Correspondingly, the original state transition matrix and process noise covariance matrix also need to be augmented in dimension. The nxn submatrix in the upper left corner remains the original matrix. The 2x2 submatrix in the lower right corner is used to describe the dynamic characteristics of the two new positioning bias states. In a preferred embodiment, the two bias states are modeled as first-order Gauss-Markov processes. Therefore, in the lower right corner of the augmented state transition matrix, the diagonal elements are exp(-positioning epoch time interval / positioning bias correlation time parameter). In the lower right corner of the augmented process noise covariance matrix, the diagonal elements are the process noise variances that describe the random walk strength of the two bias states over time.

[0118] Filter update step based on pseudo-observation:

[0119] In the standard update step of the extended Kalman filter, in addition to using traditional pseudo-range and carrier phase real observations to update the conventional state quantities, the present application independently performs an update step specifically for the augmented positioning bias states in parallel.

[0120] Step 1: Construct pseudo-observation. The obtained skewness moment pseudo-observation vector is used as the pseudo-observation for this update.

[0121] Second step: Construct the pseudo-observation model. Apply the (asymmetric-biased linear mapping pseudo-observation model) to construct the observation function and the observation Jacobian matrix required for this update.

[0122] The asymmetric-biased linear mapping pseudo-observation model constructs a brand new pseudo-observation equation, whose physical logic lies in: a persistent, eastward pointing systematic positioning bias, whose physical root must be from a strong reflector on the west side of the vehicle, and the reflector on the west side must lead to a positive bias moment pointing east. Therefore, the expected relationship between the positioning bias and the bias moment is opposite in direction. The pseudo-observation function of this model is defined as a linear transformation that multiplies the eastward and northward systematic positioning biases in the augmented state vector by a negative mapping scale factor, respectively. Its corresponding pseudo-observation Jacobian matrix is a 2x(n+2) dimensional sparse matrix, most of whose elements are zero, and only on the columns corresponding to the two positioning bias states have non-zero values - the mapping scale factor.

[0123] Third step: Construct the adaptive pseudo-observation noise covariance matrix. To characterize the assessment of the credibility of the bias moment pseudo-observation vector itself calculated at the current time, an adaptive pseudo-observation noise covariance matrix needs to be constructed for it.

[0124] Geometric quality factor coupled adaptive confidence matrix: the pseudo-observation noise covariance matrix is determined online by a geometric quality factor coupled adaptive confidence model. This model sets the matrix as a 2x2 diagonal matrix, whose diagonal elements (i.e. the pseudo-observation noise variance) are proportional to the value of the geometric dilution of precision (GDOP) of the current satellite constellation, and inversely proportional to the number of currently active visible satellites. Through this model, it can be ensured that when the satellite geometry is poor (the GDOP value is large) or the number of visible satellites is small, the value of the pseudo-observation noise variance will automatically increase, thereby making the Kalman filter automatically reduce the degree of trust in the current bias moment pseudo-observation, ensuring the robustness and stability of the entire system under various conditions.

[0125] Fourth step: Perform standard Kalman update. Using the pseudo-observation, pseudo-observation Jacobian matrix and adaptive pseudo-observation noise covariance matrix constructed above, the optimal estimation of the eastward and northward systematic positioning biases is completed by performing the standard Kalman update formula (i.e. calculating the Kalman gain, calculating the innovation, updating the state vector, updating the state covariance matrix).

[0126] After finishing all the updating steps of the current positioning epoch, obtaining the final posterior augmented state vector which contains the optimal estimation of the positioning bias. Before outputting the final positioning result, the system performs a final position compensation operation. The operation is specifically extracting the optimal estimated two-dimensional position vector from the posterior augmented state vector, and directly subtracting the two-dimensional positioning bias vector (i.e. [east-west systematic positioning bias, north-south systematic positioning bias]) also estimated in the vector, to obtain a compensated final output high-precision positioning result to the upper layer application.

[0127] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details and limit the application to the specific embodiments. Obviously, many modifications and variations can be made according to the content of the specification. The specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A method for evaluating data quality of Beidou, characterized in that, The method comprises the following steps: obtaining raw observation data of a navigation receiver for at least one Beidou satellite in a plurality of continuous positioning epochs; processing the raw observation data by using a main positioning filter to calculate post-verification residual errors of the at least one Beidou satellite in the plurality of continuous positioning epochs, thereby forming a post-verification residual error sequence; performing asymmetric quantization processing on the post-verification residual error sequence to obtain a single-satellite noise skewness coefficient value for representing a skewness degree of statistical distribution of the post-verification residual error sequence; based on the single-satellite noise skewness coefficient value of the at least one Beidou satellite and a geometric position of the at least one Beidou satellite in the sky, performing spatialized fusion processing to extract a skewness torque pseudo-observation quantity capable of macroscopically representing an asymmetry of an electromagnetic environment around the navigation receiver; utilizing the skewness torque pseudo-observation quantity to estimate and compensate for a systematic positioning bias caused by the asymmetry of the electromagnetic environment in an augmented state filtering framework, thereby outputting a final positioning result.

2. The method for evaluating the data quality of Beidou according to claim 1, wherein, The step of performing asymmetric quantization processing on the post-verification residual error sequence specifically comprises: maintaining a sliding time domain window with a preset queue length for the post-verification residual error sequence; calculating a sample mean and an unbiased sample standard deviation of all post-verification residual values in the sliding time domain window at each positioning epoch; calculating a third central moment of all post-verification residual values in the sliding time domain window based on the sample mean; calculating the single-satellite noise skewness coefficient value by using the third central moment and the unbiased sample standard deviation and through a sample quantity corrected standardization transformation.

3. The method for evaluating the data quality of Beidou according to claim 1, wherein, The step of performing spatialized fusion processing to extract the skewness torque pseudo-observation quantity specifically comprises: constructing a dynamically updated sky skewness map for the single-satellite noise skewness coefficient values of all visible Beidou satellites and corresponding geometric positions of the Beidou satellites; for each Beidou satellite in the sky skewness map, decomposing the single-satellite noise skewness coefficient value in scalar form into an east-west direction skewness component and a south-north direction skewness component according to azimuth information of the Beidou satellite; calculating a contribution weight for each Beidou satellite in the sky skewness map according to elevation information of the Beidou satellite; performing weighted averaging on the east-west direction skewness components of all Beidou satellites to obtain an east-west direction torque component of the skewness torque pseudo-observation quantity; performing weighted averaging on the south-north direction skewness components of all Beidou satellites to obtain a south-north direction torque component of the skewness torque pseudo-observation quantity.

4. The method for evaluating the data quality of Beidou according to claim 3, wherein, The contribution weight is determined by a saturation weighting model of an elevation angle inverse function, and an operation relationship of the model is defined as: taking a sine value of the elevation angle of each Beidou satellite, adding a preset saturation stabilizing factor, and then taking an inverse.

5. The method for assessing the quality of Beidou data according to claim 1, wherein, In the main positioning filter, a positioning bias state vector containing an east-west direction systematic positioning bias and a south-north direction systematic positioning bias is augmented. establishing a mathematical relationship between the east and north direction moment components of the skewness moment pseudo-observation and the east and north direction systematic positioning biases in the positioning bias state vector by constructing an asymmetric-biased linear mapping pseudo-observation model; updating the positioning bias state vector by using the asymmetric-biased linear mapping pseudo-observation model in the updating step of the main positioning filter to obtain an optimal estimation value of the positioning bias state vector; correcting the position state of the navigation receiver calculated by the main positioning filter by using the optimal estimation value of the positioning bias state vector.

6. The method for assessing the quality of Beidou data according to claim 5, wherein, In the asymmetric-biased linear mapping pseudo-observation model, the east direction moment component is proportional to the negative value of the east direction systematic positioning bias, and the north direction moment component is proportional to the negative value of the north direction systematic positioning bias.

7. The method for assessing the quality of Beidou data according to claim 5, wherein, The step of updating by using the asymmetric-biased linear mapping pseudo-observation model further comprises: constructing a pseudo-observation noise covariance matrix for representing the self-confidence of the skewness moment pseudo-observation; the pseudo-observation noise covariance matrix is determined online by a geometric quality factor coupled adaptive confidence model, which makes the element value of the pseudo-observation noise covariance matrix proportional to the geometric dilution of precision of the current satellite constellation and inversely proportional to the number of currently visible Beidou satellites.

8. A system for assessing the quality of Beidou data, characterized in that it comprises: The system is used to perform the Beidou data quality evaluation method of any one of claims 1-7, and the system comprises: a data acquisition and processing unit configured to acquire raw observation data of at least one Beidou satellite for a navigation receiver at a plurality of consecutive positioning epochs, and process the raw observation data by using a main positioning filter to calculate post-filtered residuals of the at least one Beidou satellite at the plurality of consecutive positioning epochs, thereby forming a post-filtered residual sequence; a data quality evaluation unit connected to the data acquisition and processing unit and configured to perform asymmetric quantization processing on the post-filtered residual sequence to obtain a single-satellite noise skewness coefficient value for representing the skewness degree of the statistical distribution of the post-filtered residual sequence, and perform spatialized fusion processing based on the single-satellite noise skewness coefficient value of the at least one Beidou satellite and the geometric position of the at least one Beidou satellite in the sky to extract a skewness moment pseudo-observation capable of macroscopically representing the asymmetry of the electromagnetic environment around the navigation receiver; an application verification unit connected to the data quality evaluation unit and the data acquisition and processing unit; wherein the data quality evaluation unit is further configured to output the skewness moment pseudo-observation to the application verification unit; the application verification unit is configured to receive the skewness moment pseudo-observation, estimate and compensate for a systematic positioning bias caused by the asymmetry of the electromagnetic environment in an augmented state filter framework to output a final positioning result; and the application verification unit is further configured to feed back the estimation result of the systematic positioning bias to the main positioning filter in the data acquisition and processing unit to realize closed-loop correction.

9. The system for assessing the quality of Beidou data according to claim 8, wherein, The data quality assessment unit specifically comprises: A single-satellite residual time domain window management module, configured to maintain a sliding time domain window with a preset queue length for the posterior residual sequence; A sliding window skewness calculation module, configured to calculate the single-satellite noise skewness coefficient value in the sliding time domain window; A sky skewness map construction module, configured to construct a dynamically updated sky skewness map by using the single-satellite noise skewness coefficient values of all visible Beidou satellites and the corresponding geometric positions; A skewness moment extraction module, configured to extract the skewness moment pseudo-observation by performing weighted average on the skewness components of all Beidou satellites from the sky skewness map.

Images