Beidou and INS fused real-time high-precision time service method

Abstract

The invention relates to the technical field of satellite navigation time service, in particular to a Beidou and INS fused real-time high-precision time service method, which comprises the following steps of: generating a pseudo-range residual error and a code subtraction carrier based on a Beidou satellite signal, and generating a satellite channel fuzzy vector by applying fuzzy reasoning; spatial clustering is carried out based on the multi-dimensional features of the satellite signals, and parallel sub-filters are configured for satellite clusters; the sub-filter outputs error correction and uncertainty variance by using a satellite channel fuzzy vector and carrier dynamic information; correcting the satellite observation quantity by using the error correction quantity, and constructing a composite noise covariance matrix based on the uncertainty variance; and inputting the corrected observed quantity and the composite noise covariance matrix into a navigation solver, and extracting high-precision time information. According to the method, the propagation error of the satellite signal is subjected to layered and collaborative estimation compensation, so that the positioning time service precision of the radio navigation system in a complex dynamic environment is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of satellite navigation timing technology, and in particular to a real-time high-precision timing method integrating Beidou and INS. Background Art

[0002] Radio navigation and high-precision timing applications, such as autonomous driving, drone mapping, financial transaction time synchronization, and power grid phase measurement, place extremely stringent demands on the accuracy and reliability of time bases and spatial positioning. The time and spatial position information of radio signals obtained from GNSS (particularly the BeiDou Navigation Satellite System (BDS)) enables the construction of a time and space base for radio positioning and timing. However, because BDS operates in highly attenuated electromagnetic fields, such as in urban canyons, under trees, or at tunnel entrances, it is susceptible to multipath effects, non-line-of-sight (NLOS) transceivers, and channel obstructions, resulting in severe degradation or even failure of positioning and timing accuracy. This significantly limits the application value of BDS in complex and dynamic scenarios.

[0003] While the inertial measurement unit (IMU) can effectively output information about the carrier's motion state, existing BDS radio navigation-related fusion algorithms still face core challenges in solving analytical modeling problems such as complex time-varying observation noise characteristics, especially in eliminating multipath errors. In complex dynamic scenarios, the carrier's motion state often changes so dramatically that using a fixed noise model to characterize the statistical distribution characteristics of actual radio observation errors will result in large estimation biases, significantly reducing the filter tracking effect or even causing divergence, which greatly weakens the reliability of radio measurement results and directly affects the ultimate timing accuracy.

[0004] Existing radio navigation filter models make overly simplistic and static statistical assumptions about observation noise, making it difficult to effectively identify, model, and suppress non-Gaussian, time-varying noise, such as spatiotemporally correlated multipath errors, in real time in dynamic and complex environments. Consequently, the resulting radio timing results fail to meet high-standard application requirements in terms of accuracy, continuity, and reliability. To address this issue, the present invention proposes a real-time, high-precision timing method that integrates Beidou and INS to address these technical issues. Summary of the Invention

[0005] The object of the present invention is to provide a real-time high-precision timing method integrating BeiDou and INS to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a real-time high-precision timing method integrating BeiDou and INS, comprising:

[0007] Obtaining Beidou observation data, IMU sensor data, and the real-time carrier state at the previous moment, and calculating a carrier state prediction value; generating a pseudorange residual and a code-subtracted carrier based on the Beidou observation data and the carrier state prediction value;

[0008] Based on the pseudorange residual and the code-subtracted carrier, a fuzzy inference rule is used to generate a satellite channel ambiguity vector;

[0009] Clustering positioning satellites based on Beidou observation data to generate satellite clusters; combining the satellite channel ambiguity vectors and carrier state prediction values, using parallel sub-filters to perform collaborative corrections on each satellite in the cluster, and outputting a corrected data set; the corrected data set includes error corrections and uncertainty variances;

[0010] Based on the Beidou observation data and the corrected data set, a confidence coupling method is used to generate corrected observation values and a composite noise covariance matrix; the carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into a navigation solver to generate a real-time carrier state and extract time data.

[0011] Preferably, obtaining Beidou observation data, IMU sensor data and the real-time carrier state at the previous moment, and calculating the carrier state prediction value; generating pseudorange residuals and code subtraction carriers based on the Beidou observation data and the carrier state prediction value, the specific implementation process includes:

[0012] The Beidou observation data includes the measured pseudorange, carrier phase and signal-to-noise ratio of the Beidou satellite; the IMU sensor data is an inertial measurement value, specifically including the angular increment output by the three-axis gyroscope fixed to the carrier and the velocity increment output by the three-axis accelerometer;

[0013] Extracting an IMU sensor error estimate and real-time carrier attitude information from the real-time carrier state at the previous moment, performing compensation calculations on angular increments and velocity increments using the IMU sensor error estimate to generate inertial measurement information; and generating a carrier state prediction value using a strapdown inertial navigation update algorithm based on the real-time carrier attitude information and the inertial measurement information;

[0014] The theoretical pseudorange is calculated based on the carrier state prediction value and the satellite position, and then the difference is made with the measured pseudorange to generate a pseudorange residual; the code-minus-carrier is generated based on the measured pseudorange and the carrier phase observation.

[0015] Preferably, based on the pseudorange residual and the code-subtracted carrier, a fuzzy inference rule is used to generate a satellite channel ambiguity vector, and the specific implementation process includes:

[0016] The pseudorange residual and the code-subtracted carrier are mapped into fuzzy membership using a dynamic membership function based on carrier state changes. The fuzzy membership is input into a fuzzy inference rule library, fuzzy inference rules in the form of if-then are applied, and all activated fuzzy inference rules are synthesized to quantitatively generate a satellite channel fuzzy vector.

[0017] Preferably, the dynamic membership function based on carrier state changes is specifically:

[0018] Fuzzy processing is performed on the pseudorange residual to establish a triangular membership function with a zero value as a central peak, and the positions of the boundary points of the triangular membership function are dynamically adjusted according to the carrier speed in the inertial measurement value; wherein, when the carrier speed increases, the distance between the boundary points and the central peak is reduced;

[0019] The code-subtracted carrier is fuzzified to establish a trapezoidal membership function covering a zero-value region, wherein the trapezoidal membership function includes a fully-membered platform region and an outwardly extending transition slope region; and the position of an outer boundary point of the platform region is dynamically adjusted based on the carrier angular velocity and acceleration in the inertial measurement value; wherein the outer boundary point is moved toward the zero-value direction when the modulus of the carrier angular velocity and acceleration increases.

[0020] Preferably, positioning satellites are clustered based on Beidou observation data to generate satellite clusters, and the specific implementation process includes:

[0021] The current azimuth and altitude of each satellite are calculated based on the satellite ephemeris and carrier status prediction values. A multidimensional feature time series vector is constructed for each satellite based on the azimuth, altitude, signal-to-noise ratio, and pseudorange residuals of each satellite over the past sampling period. Based on the multidimensional feature time series vector, a spatial clustering algorithm is used to cluster the satellites to form satellite clusters.

[0022] Preferably, the satellite channel ambiguity vector and the carrier state prediction value are combined to use parallel sub-filters to perform collaborative correction on the satellites in the cluster and output a corrected data set. The specific implementation process includes:

[0023] A parallel sub-filter based on the Gauss-Markov model is set for each cluster of satellites. The Gauss-Markov model parameters in the parallel sub-filter are dynamically adjusted based on the satellite channel ambiguity vector and the carrier state prediction value. The coordinated correction of satellites in the same cluster is achieved by setting the same-cluster covariance correlation terms in the noise covariance matrix of the parallel sub-filter. The multipath error correction and uncertainty variance are obtained through the parallel sub-filter to form a correction data set.

[0024] Preferably, based on the satellite channel ambiguity vector and the carrier state prediction value, the Gauss-Markov model parameters in the parallel sub-filters are dynamically adjusted. The specific process is:

[0025] Speed information is extracted from the carrier state prediction value to construct a correlation time parameter that characterizes the speed of change of the multipath error; the satellite channel ambiguity vector is analyzed to obtain the confidence of different quality levels of the satellite channel, and the basic variance parameter is generated by weighted averaging in combination with the driving noise variance; the correlation time parameter and the basic variance parameter are used together as the core parameters of the Gauss-Markov model.

[0026] Preferably, based on the Beidou observation data and the corrected data set, a confidence coupling method is used to generate corrected observation values and a composite noise covariance matrix, and the specific implementation process includes:

[0027] Subtracting a multipath error correction from the Beidou observation value to obtain a corrected observation value;

[0028] According to the carrier state prediction value, a confidence penalty factor is generated to reflect the change amplitude of the satellite signal measurement environment; the inherent noise variance of the original Beidou observation value is calculated, and the uncertainty variance and confidence penalty factor in the corrected data set are coupled to generate a composite noise covariance matrix.

[0029] Preferably, the carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into the navigation solver to generate the real-time carrier state and extract the time data. The specific implementation process includes:

[0030] Based on the corrected observation value, carrier state prediction value and composite noise covariance matrix, the carrier state prediction value is updated and corrected by using a navigation solver to generate a real-time carrier state; the receiver clock error and clock drift are extracted from the real-time carrier state to form time data.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. This system achieves intelligent, contextualized perception of satellite channel quality, moving from hard judgment to soft quantification. It innovatively introduces a fuzzy inference system based on pseudorange residuals and code-subtracted carriers, and links its membership function to the carrier's real-time dynamics (velocity, angular velocity, and acceleration). This transforms channel quality assessment from a simple binary into a fuzzy vector containing multiple levels of confidence (excellent, fair, and poor), enabling a more refined "soft quantification" of signal quality that more closely resembles human expert experience. Especially under challenging dynamic conditions such as high-speed maneuvers, the system automatically tightens its evaluation criteria, significantly improving the accuracy and robustness of error source identification in complex environments.

[0033] 2. A new hierarchical, parallel error processing architecture, transitioning from independent correction to collaborative suppression, was constructed. By introducing a satellite spatial clustering algorithm based on multidimensional feature time series vectors, clusters of satellites with similar error characteristics were first identified. Dedicated parallel subfilters were then configured for each cluster, and by setting off-diagonal correlation terms in the process noise covariance matrix, collaborative estimation of the multipath errors of satellites within the cluster was achieved. This divide-and-conquer, internally collaborative architecture effectively leverages inter-satellite correlation information for complementary corrections, thereby accurately suppressing strongly correlated errors that are difficult for individual satellites to handle, significantly improving the modeling accuracy and suppression effectiveness of the multipath error model.

[0034] 3. A confidence coupling method is proposed to construct a composite noise covariance matrix. This method incorporates three dimensions of information: first, the inherent noise variance calculated based on signal quality; second, the inherent uncertainty variance of the error correction output by the parallel sub-filters; and third, a confidence penalty factor generated based on the intensity of the carrier's motion. This method ensures that the stochastic model input to the main navigation solver more realistically reflects the credibility of each observation at the current moment, significantly improving the stability of the entire fusion system in complex dynamic environments and the reliability of the final timing results, effectively avoiding filter divergence. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of a real-time, high-precision timing method for BeiDou and INS integration proposed in an embodiment of the present invention;

[0036] Figure 2 This is a flowchart of the parallel sub-filters proposed in the embodiment of the present invention;

[0037] Figure 3 This is a flowchart of the navigation solver proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] The present invention provides a real-time high-precision timing method integrating BeiDou and INS. The technical solution is as follows:

[0040] Example 1

[0041] This embodiment provides a specific application of a real-time, high-precision timing method that integrates Beidou and INS. A typical application scenario involves a test vehicle equipped with an autonomous driving system, operating in a densely populated "urban canyon" environment. This requires providing uninterrupted, nanosecond-level timing services for the vehicle's V2X communication module and high-precision map matching module.

[0042] The test vehicle is equipped with a multi-frequency, multi-system GNSS receiver and an automotive-grade inertial measurement unit (IMU).

[0043] Reference Figure 1 , the method comprising:

[0044] S1. Obtain Beidou observation data, IMU sensor data, and the real-time carrier state at the previous moment, and calculate the carrier state prediction value; generate pseudorange residuals and code subtraction carriers based on the Beidou observation data and the carrier state prediction value;

[0045] S2. Based on the pseudorange residual and the code-subtracted carrier, a fuzzy inference rule is used to generate a satellite channel ambiguity vector;

[0046] S3. Cluster positioning satellites based on Beidou observation data to generate satellite clusters;

[0047] S4. Combining the satellite channel ambiguity vector and the carrier state prediction value, using parallel sub-filters to perform collaborative correction on each satellite in the cluster, and outputting a correction data set; the correction data set includes an error correction amount and an uncertainty variance;

[0048] S5. Based on the Beidou observation data and the corrected data set, a confidence coupling method is used to generate corrected observation values and a composite noise covariance matrix; the carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into the navigation solver to generate real-time carrier state and extract time data.

[0049] Furthermore, Beidou observation data, IMU sensor data, and the real-time carrier state at the previous moment are obtained, and a carrier state prediction value is calculated; pseudorange residuals and code-subtracted carriers are generated based on the Beidou observation data and the carrier state prediction value. Corresponding to the above step S1, the specific process includes:

[0050] In a specific processing cycle, this example adopts an update frequency of 10 Hz, that is, a period of 0.1 seconds. The system first performs data acquisition and state prediction.

[0051] Obtain BeiDou observation data for the current epoch from the GNSS receiver, including the measured pseudoranges, carrier phases, and signal-to-noise ratios of all visible satellites The angular increment output by the three-axis gyroscope and the velocity increment output by the three-axis accelerometer fixed to the carrier in the past 0.1 second are obtained from the IMU.

[0052] The estimated IMU sensor error, i.e., the zero bias of the gyroscope and accelerometer, is extracted from the real-time carrier state at the previous moment (t-1) to compensate for the currently acquired angular increment and velocity increment to obtain more accurate inertial measurement information.

[0053] According to the real-time carrier attitude information and compensated inertial measurement information at time t-1, the current moment ( ) of the carrier state prediction value, which includes the predicted position, speed and attitude of the carrier.

[0054] The theoretical pseudorange is calculated based on the carrier state prediction and the satellite position calculated from the ephemeris. This is then subtracted from the measured pseudorange to generate the pseudorange residual. The code-minus-carrier (CMC) data is generated based on the measured pseudorange and carrier phase observations.

[0055] By using high-frequency IMU data to update the strapdown inertial navigation, high-frequency continuous dynamic benchmark and prediction information are provided for subsequent observation correction and filtering.

[0056] Step S1 provides a dynamic state prediction benchmark with higher accuracy, smaller drift, and higher frequency for subsequent GNSS signal processing and data fusion through closed-loop compensation of IMU data and precise strapdown inertial navigation solution. It is the key foundation for ensuring that the entire method ultimately achieves high-precision, high-reliability, and high-continuity timing.

[0057] Furthermore, based on the pseudorange residual and the code-subtracted carrier, a fuzzy inference rule is used to generate a satellite channel ambiguity vector, corresponding to the above step S2. The specific implementation process includes:

[0058] A dynamic membership function is established, and the pseudorange residual and code-subtracted carrier are fuzzified by the dynamic membership function to generate fuzzy feature quantities; the feature quantities are input into the fuzzy reasoning rule base in the form of IF-THEN.

[0059] Specifically, the fuzzy inference rule base of this embodiment is:

[0060] The fuzzy set of channel quality is defined as (poor, medium, excellent).

[0061] Rule 1: IF (pseudorange residual is small) AND (code-subtracted carrier is stable) THEN (channel quality is good);

[0062] Rule 2: IF (pseudorange residual is large) AND (code-subtracted carrier is slow drift) THEN (channel quality is poor);

[0063] Rule 3: IF (pseudorange residual is medium) AND (code-subtracted carrier is jitter) THEN (channel quality is poor);

[0064] Rule 4: IF (pseudorange residual is small) AND (code-subtracted carrier is jittery) THEN (channel quality is medium);

[0065] Rule 5: IF (pseudorange residual is large) AND (code-subtracted carrier is stable) THEN (channel quality is medium).

[0066] Combining all activated rules, a satellite channel ambiguity vector is finally generated for each satellite, where the satellite channel ambiguity vector is [excellent: 0.1, medium: 0.8, poor: 0.1].

[0067] Fuzzy reasoning rules are introduced to evaluate satellite channel quality. Compared with the traditional hard threshold decision method based on signal-to-noise ratio and pseudorange residual, fuzzy reasoning is a soft decision that can quantify signal quality more precisely, more intelligently, and closer to human expert experience, avoiding misjudgments and missed judgments caused by improper threshold setting.

[0068] Specifically, the dynamic membership function is:

[0069] The fuzzy set of pseudorange residuals is defined as (small, medium, large).

[0070] A triangular membership function with a peak value centered at zero and descending linearly toward either side is constructed to describe the pseudorange residual. The triangle is defined by three points: the left boundary point a, the central peak point b = 0, and the right boundary point c. The positions of boundary points a and c are dynamically adjusted in real time based on the carrier velocity v extracted from the carrier state prediction.

[0071] ,

[0072] ,

[0073] in, is a base width constant, set to 2.0 in this example; is the speed influence coefficient, which is set to 0.9 in this example. When it increases, the boundary points c and a will shrink toward the central zero point, making the entire triangle membership function narrower and sharper.

[0074] The fuzzy set of code-subtracted carrier is defined as (jitter, stability, slow drift).

[0075] A trapezoidal membership function covering the zero value area is established to describe the code-subtracted carrier. The trapezoid is defined by four points: the left outer boundary point a, the left inner platform point b, the right inner platform point c, and the right outer boundary point d. Among them, the interval from b to c is completely subordinate to the stable platform area. The outer boundary points b and c of the platform area of the function are dynamically adjusted in real time according to the severity of the carrier motion state extracted from the carrier state prediction value. The severity can be determined by the acceleration and angular velocity The modulus value is used to jointly characterize.

[0076]

[0077] in, is the width of the base platform, which is set to 0.5 in this example; and are the corresponding dynamic influence coefficients, which are set to 1.0 and 0.8 respectively in this example. When the modulus of the carrier's angular velocity and acceleration increases, the platform area boundary points b and c will shrink toward the center zero point, narrowing the stable recognition range.

[0078] The membership function of fuzzy reasoning has dynamic adaptability. It links the evaluation of signal quality with the real-time motion state (speed, angular velocity, acceleration) of the carrier, so that the system can automatically tighten the evaluation criteria under harsh dynamic conditions such as high speed and maneuverability, thereby improving the model's sensitivity and adaptability to changes in the real physical environment.

[0079] Cluster positioning satellites based on BeiDou observation data to generate satellite clusters. Corresponding to the above step S3, the specific implementation process is as follows:

[0080] For each satellite, its azimuth, elevation, signal-to-noise ratio, and pseudorange residuals are collected in the past 1 second, i.e., 10 sampling periods, to form a multi-dimensional feature time series vector.

[0081] The dynamic time warping (DTW) algorithm is used to calculate the similarity distance between any two multidimensional feature time series vectors to identify satellites with similar signal behavior patterns.

[0082] Satellite A's characteristic analysis sequence in the past 1 second ; Satellite B's characteristic analysis sequence in the past 1 second . Create a 10x10 cost matrix D, each element in the matrix , represents the sequence Feature processing at the i-th moment , and sequence Feature processing at the jth moment The distance between them.

[0083] Create a 10x10 computational cost matrix Y, each element in the matrix The calculation follows a gradient rule: it is equal to the local cost of the current point , plus the minimum cumulative cost of its three adjacent predecessor nodes to the left, bottom, and left bottom. Start by gradually filling up the entire cost matrix. When the total cost matrix Y is calculated, the element in the upper right corner is The value of is the sequence and The final DTW similarity distance between them.

[0084] The above DTW calculation process is repeated for all visible satellite pairs to generate a distance matrix containing all satellite distances between each two DTW.

[0085] This paper constructs a multidimensional feature time series vector for each satellite and creatively incorporates the DTW algorithm. This allows similarity determination to move beyond the numerical value at a single point in time and instead focuses on the evolutionary trends of signal behavior patterns over time. This process-oriented analysis can more deeply and comprehensively reveal the inherent correlations of satellite signals.

[0086] The distance matrix calculated by DTW is then input into the density-based spatial clustering algorithm (DBSCAN) for clustering. In this embodiment, the parameters of DBSCAN are set to the neighborhood radius. is 0.4, the minimum number of neighbors of the core point is 2.

[0087] A satellite clustering method based on multidimensional feature time series vectors was proposed, enabling the system to identify and group satellites with similar error characteristics, especially those affected by the same local environment. This laid the foundation for subsequent targeted and efficient collaborative corrections.

[0088] Reference Figure 2 , combining the satellite channel ambiguity vector and the carrier state prediction value, using parallel sub-filters to perform collaborative correction on each satellite in the cluster, and outputting a corrected data set, corresponding to the above step S4, the specific implementation process is:

[0089] For each satellite cluster generated in the previous step, configure a parallel sub-filter based on the Gauss-Markov model.

[0090] Extract the velocity modulus from the carrier state prediction value and the angular velocity vector , combined into a comprehensive dynamic factor through coupling functions.

[0091]

[0092] in, and The weight coefficients are set to 0.5 and 2.0 in this example; is the equivalent radius, which is set to 1.5 in this example. It is used to convert the effect of angular velocity into an equivalent change in linear velocity, thereby unifying the dimensions.

[0093] Substitute this comprehensive kinetic factor D into the inverse model

[0094]

[0095] Where C is the base correlation time constant, which is set to 20 in this example.

[0096] The obtained correlation time parameter τ is used to characterize the changing speed of the multipath error.

[0097] Analyze the satellite channel ambiguity vector. In this example, the satellite channel ambiguity vector of one of the satellites is [excellent: 0.1, medium: 0.8, poor: 0.1]. Combined with the inherent noise variance of each quality level, which is set to (excellent = 0.01, medium = 0.2, poor = 1.0) in this example, the weighted average noise variance is 0.261.

[0098] Calculate the fuzzy entropy of the satellite,

[0099] ,

[0100] Since the classification is relatively certain, the entropy value is low. It is 0.922.

[0101] Set risk penalty factor for fuzzy entropy , calculate the final basic variance parameter

[0102]

[0103] Multipath error is not static; its speed of change is closely related to the dynamic environment the carrier is in. By designing the correlation time as an inverse function of the comprehensive dynamic factor, the parallel sub-filters can more effectively track and suppress time-varying multipath error in dynamic scenarios.

[0104] Fuzzy entropy is introduced to quantify the uncertainty of this expected value. This effectively prevents the filter from becoming overconfident when receiving ambiguous quality assessment information, thereby avoiding degradation or even divergence of filtering performance due to incorrect model parameters, and improving the stability and reliability of the entire fusion system.

[0105] The noise covariance matrix of the construction process is constructed. In this embodiment, a satellite cluster including three satellites A, B, and C is used.

[0106] The state vector in the parallel sub-filter is

[0107]

[0108] in, represents the multipath error of satellite A, represents the multipath error of satellite B, represents the multipath error of satellite C.

[0109] The corresponding 3x3 size process noise covariance matrix describes the variance and covariance of these error changes at each time step. The diagonal elements represents the process noise variance of the multipath error for each satellite. The off-diagonal elements It represents the covariance between the process noise of two different satellites, indicating that the random error variation of satellite A is statistically related to the random error variation of satellite B.

[0110] ,

[0111] in, The mutual correlation coefficient is set to 0.7 in this embodiment; represents the process noise variance of the satellite multipath error.

[0112] The final process noise covariance matrix in this embodiment is:

[0113]

[0114] According to the process noise covariance matrix, the coordinated correction of satellites in the satellite cluster is realized through iterative update of parallel sub-filters, and the final output is a corrected data set containing multipath error correction and uncertainty variance.

[0115] The final error correction amount in this embodiment is , the uncertainty variance is .

[0116] A parallel sub-filter and intra-cluster collaborative correction mechanism were designed. This architecture not only reduces the computational burden of the main filter (by offloading the complex observation error processing), but more importantly, by establishing covariance correlation terms within the cluster, it enables information sharing and complementarity among satellites within the cluster. This effectively corrects strongly correlated errors that are difficult for a single satellite to handle, improving overall error suppression.

[0117] The multipath model parameters in the parallel sub-filters are made dynamic. By combining carrier velocity and fuzzy inference channel quality assessment results, the core parameters of the Gauss-Markov model are dynamically adjusted. This makes the multipath error model no longer static and empirical, but a highly contextual model that reflects the current dynamics and channel environment in real time, greatly improving the accuracy of multipath error estimation.

[0118] Reference Figure 3 Based on the BeiDou observation data and the corrected data set, the confidence coupling method is used to generate the corrected observation values and the composite noise covariance matrix. Corresponding to the above step S5, the specific implementation process is as follows: The original pseudorange from the BeiDou observation data is obtained and the multipath error correction provided by the parallel sub-filter is subtracted to obtain the corrected observation value.

[0119] According to the acceleration and angular velocity modulus values in the carrier state prediction value, the penalty factor function is used:

[0120]

[0121] in, and It is a preset weight coefficient used to adjust the contribution ratio of acceleration and angular velocity. In this example, , ; is the hyperbolic tangent function; and is the threshold, i.e. the saturation adjustment parameter. In this example, , .

[0122] This factor reflects the degree of deterioration of the overall measurement environment caused by the violent motion of the carrier and is a penalty uniformly imposed on all satellite observations.

[0123] Directly linking the carrier's physical motion state with the credibility assessment of satellite observation data can reflect the deterioration of the measurement environment caused by the carrier's violent maneuvers in real time and quantitatively, allowing the random model in the navigation solver to more realistically reflect the instantaneous changes in the physical world.

[0124] The composite noise variance is generated based on the inherent noise variance and uncertainty variance, combined with the confidence penalty factor.

[0125]

[0126] in, is the composite noise variance of satellite C20; is the inherent noise variance; is the uncertainty variance; is the confidence penalty factor.

[0127] This operation is performed for all satellites to construct the composite noise covariance matrix of the navigation solver, whose diagonal elements are the composite noise variances of each satellite.

[0128] This paper proposes a confidence coupling method to generate the final composite noise covariance matrix. This method cleverly integrates the original observation noise, corrected residual uncertainty, and a penalty factor based on carrier dynamics, achieving a refined, intelligent, and robust allocation of the final weights of BDS observation information. This ensures that the navigation solver obtains the stochastic model that best reflects the current reality under all circumstances, playing a key role in preventing filter divergence and ensuring the ultimate stability of the system.

[0129] The carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into the navigation solver to generate the real-time carrier state and extract the time data. The specific implementation process corresponds to the above step S5:

[0130] The navigation solver is built based on an extended Kalman filter, whose state vector includes position error, velocity error, attitude error, IMU sensor error, and receiver clock error and drift;

[0131] Subtracting the corrected observation value from the carrier state prediction value to generate an innovation vector; calculating and generating a Kalman gain based on the carrier state prediction value and the composite noise covariance matrix;

[0132] According to the new information vector and the Kalman gain, the carrier state prediction value is corrected and updated using a main navigation filter based on the extended Kalman filter to generate a real-time carrier state; the receiver clock error and clock drift in the real-time carrier state are extracted to form time data.

[0133] Step S5 defines the path for generating the final high-precision time data, ensuring that all complex corrections and optimization processes at the front end of the entire system are processed. The final result can be seamlessly applied to the navigation solver and accurately and reliably extract the core time synchronization information, namely the clock error and drift, from the optimal state estimate, completing the closed loop of the entire method.

[0134] This paper proposes a new, systematic BeiDou / INS fusion timing framework. By incorporating innovative elements such as fuzzy reasoning, satellite clustering, parallel collaborative correction, and confidence coupling, this framework fundamentally enhances the ability to perceive, model, and mitigate satellite observation errors in complex environments, significantly improving the accuracy, continuity, and robustness of real-time, high-precision timing.

[0135] Example 2

[0136] The second embodiment of the present invention specifically involves obtaining Beidou observation data, IMU sensor data and the real-time carrier status at the previous moment, and calculating the carrier status prediction value; and generating the pseudorange residual and code subtraction carrier based on the Beidou observation data and the carrier status prediction value.

[0137] The specific steps of this embodiment are as follows: At this moment, execute as follows:

[0138] The system controller obtains the data required for the current solution cycle from the vehicle sensor network and memory. A set of observation data is collected at all times. Taking the visible Beidou C20 satellite as an example, due to the reflection of the high-rise building on the side of the vehicle, the satellite signal is subject to obvious multipath interference. Its observation data is:

[0139] ;

[0140] ;

[0141] .

[0142] The IMU original incremental data is:

[0143] ;

[0144] .

[0145] Retrieved from the navigation computer The real-time carrier state vector at time t is the optimal estimation result of the fusion filter in the previous cycle. IMU sensor error estimation at time:

[0146] ;

[0147] ;

[0148] .

[0149] use The IMU error estimated at each moment is used to compensate the newly acquired raw incremental data online to generate the corrected inertial measurement information.

[0150] ;

[0151] .

[0152] by The carrier state at the moment is taken as the starting point, and the inertial measurement information after compensation is applied to execute the strapdown inertial navigation mechanics arrangement equation to deduce The carrier state prediction value at time .

[0153] use The vehicle's predicted position at the time and the C20 satellite calculated from the ephemeris are The spatial position at the moment is used to calculate the theoretical geometric distance between the two. Using the Saastamoinen model and ephemeris parameters, the tropospheric delay, ionospheric delay, and satellite clock error are calculated and compensated. The theoretical pseudorange obtained by calculation .

[0154] Subtract the measured pseudorange from the theoretical pseudorange to obtain the pseudorange residual that reflects the current observation quality. , revealing that the C20 satellite signal is currently suffering from severe multipath pollution.

[0155] The code-minus-carrier (CMC) calculation formula is:

[0156] ,

[0157] Where λ is the wavelength of the C20 satellite signal, which is 0.1903 meters.

[0158] The absolute value of the CMC is the key basis for judging the intensity and dynamic characteristics of the multipath effect.

[0159] Example 3

[0160] To further clarify the objectives, technical solutions, and advantages of the present invention, the following will further describe the technical solutions of the present invention in detail and in detail with reference to specific embodiments. This embodiment specifically relates to a technical solution for generating satellite channel ambiguity vectors using fuzzy inference rules based on the pseudorange residual and code-subtracted carrier. This embodiment continues the application scenario of Example 2.

[0161] Pseudorange residuals

[0162] Code-subtraction carrier variation

[0163] Dynamic information extracted from the vehicle state prediction value: vehicle speed modulus

[0164] Vehicle angular velocity modulus

[0165] Vehicle acceleration modulus

[0166] The system fuzzifies the pseudorange residuals and code-subtraction carrier variations and maps them to memberships of different fuzzy sets. Three fuzzy sets are defined for the pseudorange residuals: negative large (NB), zero (ZO), and positive large (PB). The membership function of the zero (ZO) set adopts a triangular membership function with a peak centered at zero. The boundary points of the triangular membership function are The position of the vehicle is determined by the vehicle speed. Dynamic adjustment. The adjustment rule is to reduce the distance between the boundary point and the central peak when the carrier speed increases, that is, to tighten the interval determined as zero and improve the strictness of the judgment. The dynamic adjustment formula is designed as follows:

[0167] ,

[0168] ,

[0169] in, is the basic boundary value in the stable state, which is set to rice. is the speed influence coefficient, which is set as =0.1.

[0170] According to the current speed , calculate the dynamic boundary point to be 1.2 meters

[0171] This means that at the current At a speed of When the probability is zero, it is considered to be zero.

[0172] The input pseudorange residual =−4.58 m, substituted into the membership function after dynamic adjustment.

[0173] is the code-to-carrier variation Two fuzzy sets are defined: stable and fluctuating. The stable set uses a trapezoidal membership function that covers the zero value region.

[0174] The outer boundary points of the complete membership platform area of the trapezoidal membership function are based on the angular velocity of the vehicle. and acceleration Dynamic adjustment is performed. The rule is: when the angular velocity and acceleration modulus increase, the outer boundary point is moved toward the zero value direction, that is, the stable platform area is tightened. The specific dynamic adjustment formula can be designed as:

[0175]

[0176] in, is the basic boundary value under static conditions. In this example, rice. and are the influence coefficients of angular velocity and acceleration respectively. In this example, they are set as , .

[0177] According to the current vehicle dynamics , , the boundary of the dynamic platform area is calculated to be 0.28 meters.

[0178] Subtract the input code from the carrier change =0.15 m, and substitute it into the membership function after dynamic adjustment.

[0179] The calculated fuzzy membership degrees are input into the fuzzy inference rule base to generate the final satellite channel fuzzy vector.

[0180] The fuzzy inference rule base includes an expert knowledge-based IF-THEN rule base for evaluating satellite channel quality. Channel quality can be categorized as Good, Medium, or Poor. The rule base is as follows:

[0181] Rule 1: IF (pseudorange residual is zero) AND (CMC variation is stable) THEN (channel quality is excellent);

[0182] Rule 2: IF (pseudorange residual is zero) AND (CMC variation is fluctuating) THEN (channel quality is medium);

[0183] Rule 3: IF (pseudorange residual is negative and large) AND (CMC variation is stable) THEN (channel quality is poor);

[0184] Rule 4: IF (pseudorange residual is positive) AND (CMC variation is stable) THEN (channel quality is poor);

[0185] Rule 5: IF (pseudorange residual is negative and large) AND (CMC variation is fluctuating) THEN (channel quality is poor);

[0186] Rule 6: IF (pseudorange residual is positive) AND (CMC variation is fluctuating) THEN (channel quality is poor).

[0187] When the membership of the C20 satellite is input into the rule base, the pseudorange residual is significantly negative and the CMC variation is stable.

[0188] The activation strength of rule 1 is 0, and its contribution to the final conclusion of "good" is zero.

[0189] The activation strength of Rule 2 is 1. This rule is fully activated and its confidence in concluding that the channel quality is poor is 1.0.

[0190] Combining all activated rules, defuzzification is performed through weighted average method, and finally a vector representing the quality of C20 satellite channel is quantified. Based on the above reasoning, the fuzzy vector of C20 satellite is: =[Confidence (excellent), Confidence (medium), Confidence (poor)] = [0.0, 0.1, 0.9].

[0191] Here, [0.0, 0.1, 0.9] means that there is a 90% confidence that the channel quality of the C20 satellite is poor, a 10% confidence that it is medium, and a 0% confidence that it is excellent.

[0192] Example 4

[0193] This embodiment specifically describes the technical solution process of clustering positioning satellites, using parallel sub-filters to perform collaborative correction on satellites in the cluster, and dynamically adjusting Gauss-Markov model parameters in the parallel sub-filters.

[0194] In complex environments like urban canyons, satellite signals from different directions experience significantly different interference characteristics. Signals from multiple satellites on the same narrow street may be simultaneously reflected by the glass curtain wall of a particular building, causing their multipath errors to exhibit high correlation. Treating all satellite errors as independent events loses this valuable information, affecting the accuracy of error estimates.

[0195] This embodiment continues the scenario of an autonomous vehicle driving in a city CBD. Based on the previous embodiment, the specific process of satellite clustering, dynamic adjustment of filter parameters, and collaborative error correction is described in detail.

[0196] exist At this moment, there are 5 BeiDou satellites visible to the vehicle: C19, C20, C21, C28, and C30. Based on the satellite ephemeris and the vehicle's predicted position, the azimuth angle at the current moment is calculated. and altitude angle , extract the observation data sequence of the past 10 sampling periods to form time series features.

[0197] Specifically, the feature vector constructed for each satellite is The form is

[0198] ;

[0199] in, are the current azimuth and altitude angles, is the average signal-to-noise ratio over the past 10 periods, is the average pseudorange residual over the past 10 cycles, is the standard deviation of the pseudorange residuals over the past 10 cycles, reflecting its stability.

[0200] After system calculation and integration, the characteristic vector data of each satellite is obtained:

[0201] C19: [290°, 25°, 40dB, -2.5m, 1.8m];

[0202] C20: [110°, 35°, 38dB, -4.2m, 2.5m];

[0203] C21: [115°, 38°, 39dB, -3.9m, 2.2m];

[0204] C28: [180°, 75°, 48dB, 0.2m, 0.5m];

[0205] C30: [210°, 70°, 47dB, 0.4m, 0.6m].

[0206] The five multidimensional feature vectors are used as input to perform clustering using the DBSCAN spatial clustering algorithm. Based on the above data, the DBSCAN algorithm identifies the following three satellite clusters:

[0207] Satellite cluster A (low altitude, poor quality): {C20, C21}. These two satellites have similar azimuth and elevation angles, and their historical pseudorange residuals show significant negative offsets and large fluctuations, indicating that they may be affected by the same set of reflecting surfaces.

[0208] Satellite cluster B (high altitude, high quality): {C28, C30}. These two satellites have high elevation angles, good signal-to-noise ratios, and small and stable pseudorange residuals.

[0209] Satellite cluster C (isolated point): {C19}, which is quite different from other satellites and forms a class of its own.

[0210] An accurate error model is set for each parallel sub-filter. The present invention adopts a first-order Gauss-Markov model to describe the multipath error. The timing characteristics of the core parameter is the correlation time and the driving noise variance .

[0211] Take the C20 satellite in satellite cluster A as an example:

[0212] Extract the velocity modulus from the carrier state prediction value and the angular velocity vector , combined into a comprehensive dynamic factor through coupling functions.

[0213]

[0214] in, and The weight coefficients are set to 0.5 and 2.0 in this example; is the equivalent radius, which is set to 1.5 in this example. It is used to convert the effect of angular velocity into an equivalent change in linear velocity, thereby unifying the dimensions.

[0215] Substitute this comprehensive kinetic factor D into the inverse model

[0216]

[0217] Where C is the basic correlation time constant, set to 20 in this example. The resulting correlation time parameter, τ, characterizes the rate of change of the multipath error. In this example, the multipath error correlation time of C20 is 12.5 seconds.

[0218] The ambiguity vector of C20 satellite is , representing confidence levels of 0.0, 0.1, and 0.9 for excellent, medium, and poor channel quality, respectively. The base variance is calculated by weighting the driving noise variances corresponding to different quality levels according to the confidence level in the fuzzy vector. In this example, the default base variances for each quality level are (Excellent = 0.12, Medium = 1.02, Poor = 3.00), resulting in a weighted average noise variance of 2.802.

[0219] Calculate the fuzzy entropy of the satellite:

[0220] ,

[0221] Set risk penalty factor for fuzzy entropy , the final basic variance parameter is calculated to be 3.459.

[0222] For each satellite cluster, an independent parallel sub-filter is configured. The task of each sub-filter is to estimate the multipath errors of all satellites in the cluster.

[0223] In order to achieve the coordinated correction of satellites in the same cluster, the covariance correlation term of the same cluster is set in the process noise covariance matrix of the sub-filter.

[0224] For cluster A, the diagonal elements of satellites C20 and C21 are their respective basis variance parameters, while the correlation coefficients in the off-diagonal elements are A coupling relationship is established between the C20 and C21 error estimates. When the C20 measurement update indicates an increase in its multipath error, the filter model uses this correlation term to simultaneously improve the C21 multipath error estimate.

[0225] Run each sub-filter in parallel. Each filter takes the pseudo-range residual of the corresponding satellite as the observation input to perform state estimation.

[0226] Taking C20 as an example, after filtering, the sub-filter of cluster A outputs the multipath error estimation for the C20 satellite.

[0227] Error correction amount: This is the filter's The best estimate of the multipath component in the meter pseudorange residuals.

[0228] Uncertainty variance: This is the variance of the error correction estimate itself, which represents its credibility.

[0229] Example 5

[0230] Reference Figure 3 This embodiment specifically describes the technical solution process for generating corrected observation values and composite noise covariance matrix, generating real-time carrier status, and extracting time data.

[0231] According to the original measured pseudorange of C20 satellite =21877456.22 meters and the error correction of C20 satellite: Meters, get the corrected observation value

[0232] Calculate the inherent noise variance of C20 , calculated based on the signal-to-noise ratio and elevation angle using the standard formula for inherent noise =0.09.

[0233] According to the acceleration and angular velocity modulus values in the carrier state prediction value, a confidence penalty factor is generated through the penalty factor function:

[0234]

[0235] in, and It is a preset weight coefficient used to adjust the contribution ratio of acceleration and angular velocity. In this example, , ; is the hyperbolic tangent function; and is the threshold, i.e. the saturation adjustment parameter. In this example, , .

[0236] This factor reflects the degree of deterioration of the overall measurement environment caused by the violent movement of the carrier, and is a penalty imposed uniformly on all satellite observations. The confidence penalty factor is calculated as

[0237] The composite noise variance of the C20 satellite is generated based on the inherent noise variance and uncertainty variance, combined with the confidence penalty factor.

[0238]

[0239] in, is the composite noise variance of satellite C20; is the inherent noise variance; is the uncertainty variance; is the confidence penalty factor.

[0240] This operation is performed for all satellites to construct the composite noise covariance matrix of the navigation solver, whose diagonal elements are the composite noise variances of each satellite.

[0241] Define the state vector of the extended Kalman filter-based navigation solver, including position, velocity, attitude error, IMU sensor error, and receiver clock error and drift. Subtract the corrected observations from the vehicle state predictions to generate an innovation vector. Calculate the Kalman gain based on the vehicle state predictions and the composite noise covariance matrix.

[0242] According to the new information vector and Kalman gain, the navigation solver based on the extended Kalman filter is used to correct and update the carrier state prediction value to generate the real-time carrier state.

[0243] The navigation solver outputs the updated real-time vehicle state, which includes the vehicle's real-time position, velocity, attitude, IMU error, and the optimal estimate vector of the receiver clock error and drift.

[0244] From this state vector, directly extract the state quantities related to timing:

[0245] Receiver clock error Second;

[0246] Receiver clock drift Second.

[0247] The final output is real-time time data consisting of the receiver clock difference and the receiver clock drift.

[0248] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A real-time high-precision timing method integrating BeiDou and INS, characterized in that: include: Obtain BeiDou observation data, IMU sensor data, and the real-time carrier status at the previous moment, and calculate the carrier status prediction value; Generate a pseudorange residual and a code-subtracted carrier based on the BeiDou observation data and the carrier state prediction value; Based on the pseudorange residual and the code-subtracted carrier, a fuzzy inference rule is used to generate a satellite channel ambiguity vector; Cluster positioning satellites based on BeiDou observation data to generate satellite clusters; Combining the satellite channel ambiguity vector and the carrier state prediction value, using parallel sub-filters to perform collaborative correction on each satellite in the cluster, and outputting a corrected data set; The correction data set includes an error correction amount and an uncertainty variance; Based on the BeiDou observation data and the corrected data set, a confidence coupling method is used to generate corrected observation values and a composite noise covariance matrix; The carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into the navigation solver to generate the real-time carrier state and extract the time data.

2. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: The method comprises the following steps: obtaining Beidou observation data, IMU sensor data and the real-time carrier state at the previous moment, and calculating a carrier state prediction value; generating a pseudorange residual and a code-subtraction carrier based on the Beidou observation data and the carrier state prediction value, and the specific implementation process comprises: the Beidou observation data comprises the measured pseudorange, carrier phase and signal-to-noise ratio of the Beidou satellite; the IMU sensor data is an inertial measurement value, specifically comprising an angular increment output by a three-axis gyroscope fixed to the carrier and a velocity increment output by a three-axis accelerometer; extracting an IMU sensor error estimate and real-time carrier attitude information from the real-time carrier state at the previous moment, using the IMU sensor error estimate to perform compensation calculation on the angular increment and velocity increment to generate inertial measurement information; generating a carrier state prediction value based on the real-time carrier attitude information and the inertial measurement information using a strapdown inertial navigation update algorithm; calculating a theoretical pseudorange based on the carrier state prediction value and the satellite position, and then performing a subtraction with the measured pseudorange to generate a pseudorange residual; and generating a code-subtraction carrier based on the measured pseudorange and carrier phase observation value.

3. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: The satellite channel ambiguity vector is generated based on the pseudorange residual and the code-subtracted carrier using fuzzy inference rules. The specific implementation process includes: using a dynamic membership function based on carrier state changes to map the pseudorange residual and the code-subtracted carrier into fuzzy membership; inputting the fuzzy membership into a fuzzy inference rule library, applying fuzzy inference rules in the form of IF-THEN, and then synthesizing all activated fuzzy inference rules to quantitatively generate the satellite channel ambiguity vector.

4. The real-time high-precision timing method for integrating BeiDou and INS according to claim 3 is characterized in that: The dynamic membership function based on carrier state changes is specifically as follows: fuzzifying the pseudorange residual to establish a triangular membership function with a zero value as a central peak, and dynamically adjusting the positions of the boundary points of the triangular membership function according to the carrier velocity in the inertial measurement value; wherein, when the carrier velocity increases, the distance between the boundary point and the central peak is reduced; fuzzifying the code-subtracted carrier to establish a trapezoidal membership function covering the zero value area, wherein the trapezoidal membership function includes a fully affiliated platform area and an outward-extending transition slope area; and dynamically adjusting the position of the outer boundary points of the platform area according to the carrier angular velocity and acceleration in the inertial measurement value; wherein, when the modulus of the carrier angular velocity and acceleration increases, the outer boundary points are moved toward the zero value direction.

5. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: The method of clustering positioning satellites based on Beidou observation data to generate satellite clusters specifically includes: calculating the azimuth and altitude of each satellite at the current moment based on the satellite ephemeris and carrier state prediction value; constructing a multidimensional feature time series vector for each satellite based on the azimuth, altitude, signal-to-noise ratio and pseudorange residual of each satellite in the past sampling period; and clustering the satellites using a spatial clustering algorithm based on the multidimensional feature time series vector to form satellite clusters.

6. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: The satellite channel ambiguity vector and the carrier state prediction value are combined to use parallel sub-filters to perform collaborative correction on the satellites in the cluster and output a corrected data set. The specific implementation process includes: setting a parallel sub-filter based on a Gaussian-Markov model for each cluster of satellites, and dynamically adjusting the Gaussian-Markov model parameters in the parallel sub-filter based on the satellite channel ambiguity vector and the carrier state prediction value; realizing collaborative correction of satellites in the same cluster by setting the same-cluster covariance related terms in the process noise covariance matrix of the parallel sub-filter; and obtaining multipath error correction and uncertainty variance through the parallel sub-filter to form a corrected data set.

7. The real-time high-precision timing method for integrating BeiDou and INS according to claim 6 is characterized in that: The method dynamically adjusts the Gauss-Markov model parameters in the parallel sub-filter based on the satellite channel ambiguity vector and the carrier state prediction value. The specific process is as follows: extracting speed information from the carrier state prediction value to construct a correlation time parameter that characterizes the speed of change of the multipath error; parsing the satellite channel ambiguity vector to obtain the confidence of different quality levels of the satellite channel, and generating a basic variance parameter based on the inherent noise variance; and using the correlation time parameter and the basic variance parameter together as the core parameters of the Gauss-Markov model.

8. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: Based on the Beidou observation data and the corrected data set, a confidence coupling method is used to generate corrected observation values and a composite noise covariance matrix. The specific implementation process includes: based on the Beidou observation value, subtracting the multipath error correction amount to obtain the corrected observation value; generating a confidence penalty factor based on the carrier state prediction value to reflect the change amplitude of the satellite signal measurement environment; calculating the inherent noise variance of the original Beidou observation value, and coupling it with the uncertainty variance and confidence penalty factor in the corrected data set to generate a composite noise covariance matrix.

9. The real-time high-precision timing method for integrating BeiDou and INS according to claim 1, characterized in that: The carrier state prediction value, the corrected observation value and the composite noise covariance matrix are input into the navigation solver to generate the real-time carrier state and extract the time data. The specific implementation process includes: based on the corrected observation value, the carrier state prediction value and the composite noise covariance matrix, the carrier state prediction value is updated and corrected by the navigation solver to generate the real-time carrier state; and the receiver clock error and clock drift are extracted from the real-time carrier state to form the time data.

Images