A Beidou + 5G multi-dimensional measurement covariance weighted positioning method based on Kalman filtering

Abstract

This invention discloses a BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering. It introduces a multidimensional measurement covariance weighting mechanism, which effectively suppresses the impact of abnormal observations caused by multipath and non-line-of-sight propagation on filtering through weighting based on median absolute deviation. Considering the characteristics of 5G base station layout, it reduces the impact of adverse geometric configurations on positioning results through weighting based on horizontal geometric dilution and vertical sensitivity. For the frequent switching of satellite and base station links in urban environments, a new link weighting strategy is designed to effectively smooth transient positioning jumps that may occur when new links appear. Unlike traditional single-indicator covariance weighting, this method comprehensively considers data outlier degree, geometric distribution, and link time-varying characteristics. In an urban canyon simulation environment, compared with traditional adaptive Kalman filtering, this method reduces the root mean square error of positioning in the east, north, and sky directions, improving positioning accuracy and robustness.

Description

Technical Field

[0001] This invention belongs to the field of satellite navigation technology, specifically relating to a BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering. Background Technology

[0002] Positioning technology is a crucial foundation for achieving precise operations and reliable functioning in scenarios such as intelligent driving, low-altitude operations, and safety inspections. Therefore, researchers have conducted extensive research on high-precision positioning technologies in complex environments. In the field of multi-source fusion positioning, the synergistic application of BeiDou and 5G technologies, with their advantages such as satellite-to-ground complementarity, has become an effective way to solve positioning bottlenecks in complex scenarios such as urban canyons and indoor-outdoor switching, providing positioning systems with broader scenario adaptability and higher positioning accuracy.

[0003] However, BeiDou satellite signals are susceptible to multipath interference, while 5G positioning signals suffer from frequent link switching and are severely affected by geometric configurations, resulting in low positioning accuracy after single or simple fusion. Simultaneously, the covariance characteristics of multi-source heterogeneous data vary significantly, making it difficult for traditional fusion methods to accurately match data confidence levels, easily leading to positioning result deviations. Kalman filtering, as a classic state estimation method, is widely used in data denoising and fusion optimization; however, the fixed setting of the covariance matrix in traditional Kalman filtering cannot adaptively adjust according to the dynamic error characteristics of BeiDou + 5G data, limiting further improvements in positioning accuracy. Therefore, a multi-dimensional measurement covariance weighting scheme is urgently needed to achieve accurate matching and dynamic correction of the error characteristics of multi-source data, thereby improving the accuracy and robustness of the BeiDou + 5G fused positioning system. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a multi-dimensional measurement covariance weighted positioning method based on Kalman filtering for BeiDou + 5G. This method can solve the problem that the fixed covariance setting of traditional Kalman filtering is difficult to adapt to the dynamic error characteristics of multi-source data in complex environments by dynamically weighting the covariance matrix of BeiDou and 5G multi-source positioning data in multiple dimensions, thereby improving the accuracy and robustness of fusion positioning.

[0005] A multi-dimensional measurement covariance weighted positioning method based on Kalman filtering for BeiDou + 5G includes:

[0006] Step 1: Process the state vector of the Kalman filter Define; Step 2: In the current epoch The pseudorange observation value based on Beidou satellites The distance observation value equivalent to the arrival time of 5G base stations is Observation equations for BeiDou satellite and 5G measurement values ​​were constructed respectively; Using the state estimation of the previous epoch Covariance Matrix Combined with the state transition matrix Update to obtain the current epoch. Next, the predicted state vector and predicted covariance : (4) (5) in The noise covariance matrix of the previous time step; Construct a nonlinear observation function composed of the BeiDou pseudorange observation equation and the 5G TOA observation equation. Assume there are a total of at time k Beidou satellites and One 5G base station, The expression takes the following form: (6) in, ... The first one... the second one The distance between a BeiDou satellite and the predicted location of the user terminal. , These are the ionospheric and tropospheric delays of the first Beidou satellite, respectively. , The first The ionospheric and tropospheric delay of a BeiDou satellite; , The first and the second, respectively. The distance between a 5G base station and the predicted location of a user terminal; Represents the speed of light; express For receiver clock bias; According to formula (6), each state variable in the matrix is ​​replaced with the predicted state variable at the current time k to obtain the observed predicted value at the current epoch. ; Simultaneously, the actual collected BeiDou pseudorange and 5G TOA data are combined into a measurement vector. Calculate the innovation vector The formula is as follows: (7) Step 3: Calculate the covariance inflation factor, including the outlier factor for each observation i. Horizontal geometric factors Vertical sensitivity factor and link continuity factor ; Step 4: Based on the covariance inflation factor in Step 3, update the diagonal elements of the measurement covariance using the unified weighting formula. : (18) in, The set measurement variance; Step 5: Kalman Measurement Update Using weighted measurement covariance Calculate Kalman gain : (19) in, This is the filtered observation matrix; Correcting the state vector using Kalman gain and posterior covariance matrix : (20) (twenty one) in, Represents the identity matrix; Final output The positional components in the equation are used as the high-precision positioning result for the current epoch. Step 6: Epoch Iteration and Termination Determination Determine whether the current positioning task has ended or whether there is still observation data input for subsequent epochs: If there is still data, then set the time epoch. And estimate the posterior state at the current time. and posterior covariance matrix As prior information for the next epoch, return to step 2 for the next round of recursive calculation; if there is no data, end the current positioning.

[0007] A better outlier inflation factor The calculation process is as follows: The dispersion of the innovation is assessed using the median absolute deviation statistic; first, the standardization factor is calculated. : (8) in, This indicates taking the median of the data; Calculate normalized new information Outlier inflation factor Defined as: (9) in, Indicates an exponent; Represents the information vector of the i-th observation; This indicates the set threshold.

[0008] Preferred horizontal geometric factor The calculation process for the horizontal geometric distribution quality of 5G observations is as follows: Calculate the horizontal precision factor corresponding to the 5G observation matrix The calculation formula is as follows:

[0009] in, and The weight coefficient matrix of the 5G observation equation is shown below. The eastward and northward variance components on the diagonal. Kalman filter observation matrix The submatrix corresponding to 5G observations; Define the original scaling value : (10) in, This represents the horizontal geometric reference value for 5G base stations; To avoid drastic fluctuations in this factor, an exponential moving average is used. Smoothing process is performed to obtain : (11) in, Indicates smoothing weights; Horizontal geometric factor Defined as: (12).

[0010] Preferred vertical sensitivity factor For 5G observation, the calculation process includes: First, extract the current 5G observations into the Kalman filtered observation matrix. Submatrix corresponding to 5G observations The third element of the corresponding row is denoted as the vertical component of the gaze vector. Calculate its absolute value Meanwhile, define vertical geometric indices. : (13) in, This represents the vertical reference value for 5G base stations; Define the original vertical penalty value for: (14) in, Indicates the vertical expansion index; Vertical sensitivity factor Defined as: (15).

[0011] A better link continuity factor The calculation process is as follows: For observation i, define a smoothing variable for the current epoch k. For BeiDou satellites, if it is a newly emerging satellite link, then ,otherwise For 5G base stations, a recursive smooth update is adopted as follows: (16) in, As a new link smoothing factor, , This is the initial expansion coefficient; if it is a newly emerging 5G base station link, then... ,otherwise ; The link continuity factor is defined as: (17).

[0012] The present invention has the following beneficial effects: This invention introduces a multi-dimensional measurement covariance weighting mechanism into the Kalman filtering framework for BeiDou + 5G hybrid positioning. This method effectively suppresses the impact of anomalous observations caused by multipath and non-line-of-sight propagation on the filtering by weighting based on median absolute deviation. Considering the characteristics of 5G base station layout, it reduces the impact of adverse geometric configurations on positioning results by weighting based on horizontal geometric dilution and vertical sensitivity. For the frequent switching of satellite and base station links in urban environments, a new link weighting strategy is designed to effectively smooth out transient positioning jumps that may occur when new links appear. Unlike traditional single-indicator covariance weighting, this method comprehensively considers data outlier degree, geometric distribution, and link time-varying characteristics. Experiments show that in an urban canyon simulation environment, compared with traditional adaptive Kalman filtering, this method reduces the root mean square error of positioning in the east, north, and sky directions, improving positioning accuracy and robustness. Attached Figure Description

[0013] Figure 1 This is a flowchart of the method of the present invention.

[0014] Figure 2 This is a logical diagram of the multidimensional measurement covariance weighting module in this invention.

[0015] Figure 3 This is a two-dimensional simulation scene diagram in this invention.

[0016] Figure 4 This is a schematic diagram of the three-dimensional simulation scene in this invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] To address the issue of decreased positioning accuracy in urban canyon environments due to large fluctuations in observation quality in existing BeiDou+5G fusion positioning methods, this invention provides a BeiDou+5G multidimensional measurement covariance weighted positioning method based on Kalman filtering. This method, based on the Kalman filtering framework, focuses on dynamically adjusting the measurement noise covariance matrix through a multidimensional measurement covariance weighting module. This is short for measuring covariance. The overall steps are as follows: Step 1: Parameter Initialization Define the state variables of the Kalman filter. State vector. Includes the current epoch The parameters such as user location, speed, and receiver clock error at any given time are specifically expressed as follows: (1) in, For three-dimensional position coordinates, For three-dimensional velocity, This refers to the receiver clock bias. Simultaneously, the four core weighting factors and their parameters of the multidimensional measurement covariance weighting module are initialized, following these steps: Step 1-1: Initialization of Outlier Inflation Factor Parameters Threshold for determining Median Absolute Deviation (MAD) ,index Perform initialization.

[0019] Step 1-2: Initialization of horizontal geometric factor parameters 5G horizontal geometric reference value Smoothing weights Perform initialization.

[0020] Steps 1-3: Initialization of Vertical Sensitivity Factor Parameters 5G vertical reference value Vertical expansion index Perform initialization.

[0021] Steps 1-4: Initialization of Link Continuity Factor Parameters Smoothing factor for new links Initial expansion coefficient , Perform initialization.

[0022] Steps 1-5: Initialization of Kalman filter parameters For the state covariance matrix process noise covariance matrix Initialize the parameters such as the set measurement variance.

[0023] Step 2: Construct a nonlinear observation function composed of the BeiDou pseudorange observation equation and the 5G TOA observation equation: In the current epoch The pseudorange observation value of Beidou satellite The equivalent distance observation value for 5G Time of Arrival (TOA) is The observation equations for BeiDou and 5G measurements are modeled as follows: (2) (3) in, , These represent the actual distances between the BeiDou satellite and the 5G base station and the user terminal at the current time k. For receiver clock bias, this embodiment assumes that BeiDou and 5G systems share a clock reference. and These are the ionospheric and tropospheric delays, respectively. At the speed of light, , These are the measurement noises for BeiDou and 5G, respectively.

[0024] Using the state estimation of the previous epoch Covariance Matrix Combined with the state transition matrix Update to obtain the current epoch. Next, the predicted state vector and predicted covariance : (4) (5) in Let be the process noise covariance matrix of the previous time step.

[0025] Construct a nonlinear observation function composed of the BeiDou pseudorange observation equation and the 5G TOA observation equation. Assume there are a total of at time k Beidou satellites and One 5G base station, The expression takes the following form: (6) in, ... The first one... the second one The distance between a BeiDou satellite and the predicted location of the user terminal. , For the ionospheric and tropospheric delay of the first Beidou satellite, , For the first The ionospheric and tropospheric delay of a Beidou satellite. , The first and the second are respectively. The distance between a 5G base station and the predicted location of a user terminal.

[0026] According to formula (6), each state variable in the matrix is ​​replaced with the predicted state variable at the current time k to obtain the observed predicted value at the current epoch. .

[0027] Simultaneously, the BeiDou pseudorange and 5G TOA data actually collected in step 2 are combined into a measurement vector. Calculate the innovation vector The formula is as follows: (7) Step 3: Calculation of covariance inflation factor Prior to the Kalman update, this invention defines four independent covariance inflation factors for multidimensional measurement covariance weighting: outlier factor Horizontal geometric factors Vertical sensitivity factor and link continuity factor For each observation Calculate each of the four factors separately, and then calculate the final measurement variance using a unified weighting formula. Finally, the corrected variance of all observations is used. Construct the final adaptive measurement covariance The logic diagram is as follows: Figure 2 As shown. The steps for calculating the covariance inflation factor are as follows: Step 3-1: Calculate the outlier inflation factor

[0028] To combat gross errors, the median absolute deviation statistic is used to assess the dispersion of the innovation. First, the standardization factor is calculated. : (8) in,

[0029] Calculate normalized new information Outlier inflation factor Defined as: (9) in, Let represent the information vector of the i-th observation.

[0030] This factor constructs an adaptive suppression mechanism for outliers. For observations that have significant biases due to non-line-of-sight propagation or multipath effects, this mechanism rapidly amplifies the corresponding noise variance using the properties of an exponential function, thereby reducing the weight of outlier data and ensuring the robustness of the system.

[0031] Step 3-2: Calculate the horizontal geometric factor

[0032] This factor is primarily for 5G observations and is used to reflect the quality of the horizontal geometric distribution of current observations.

[0033] Calculate the horizontal precision factor corresponding to the 5G observation matrix The calculation formula is as follows:

[0034] in, and The weight coefficient matrix of the 5G observation equation is shown below. The eastward and northward variance components on the diagonal. Kalman filter observation matrix The submatrix corresponding to 5G observations.

[0035] Define the original scaling value : (10) To avoid drastic fluctuations in this factor, an exponential moving average is used. Smoothing process is performed to obtain : (11) Horizontal geometric factor Defined as: (12) This factor is used to quantify the horizontal geometric distribution quality of 5G observations. By calculating the ratio of the current horizontal accuracy factor to the reference value and performing smoothing, the variance of 5G observation noise is dynamically adjusted: when the geometric configuration is poor, the variance is amplified to reduce the confidence level, and vice versa, the confidence level is maintained or increased, thereby achieving adaptive weighting of data with different geometric quality.

[0036] Step 3-3: Calculate the vertical sensitivity factor

[0037] This factor is primarily designed for 5G observations and is used to address the issue of weak vertical observation geometry in urban environments, where 5G base stations are often deployed on roadsides or rooftops, resulting in signal loss at high elevation angles.

[0038] First, extract current 5G observations. The third element of the corresponding row is denoted as the vertical component of the gaze vector. Calculate its absolute value. . The smaller the value, the more horizontal the observation link tends to be, resulting in a smaller contribution to the vertical position calculation and a greater likelihood of introducing elevation errors. Simultaneously, a vertical geometric index is defined. : (13) Define the original vertical penalty value for: (14) Based on this indicator, the vertical sensitivity factor Defined as: (15) This factor reduces the weight of poorly shaped links in the vertical state update by imposing a non-linear variance penalty on base station links with low elevation angles or those near the user's horizontal plane. This effectively suppresses positioning divergence in the elevation direction and improves the overall stability of 3D positioning.

[0039] Steps 3-4: Calculate the link continuity factor

[0040] This factor addresses the issue of abrupt changes in the visibility of BeiDou satellites and 5G base stations caused by user movement in urban canyon environments. It is used to suppress the jumps in positioning results caused by the new access link not yet converging, aiming to smooth the transient errors generated during the switching between old and new links and ensure the continuity of the positioning trajectory.

[0041] To monitor link status, a smoothed variable is defined for observation i at time k. For BeiDou satellites, if it is a newly emerging satellite link, then ,otherwise For 5G base stations, a recursive smooth update is adopted as follows: (16) If it is a newly emerging 5G base station link, then ,otherwise .

[0042] The link continuity factor is defined as: (17) This factor employs differentiated suppression strategies based on the type of observation source: for newly emerging BeiDou satellite links, an instantaneous variance penalty is applied; for newly connected 5G base station links, a penalty measure with recursive attenuation characteristics is applied. Through this hierarchical processing mechanism, the confidence weight of various signals that have not yet converged in the initial stage of access is reduced, thereby effectively suppressing transient positioning abrupt changes caused by drastic changes in network topology and improving the continuity and smoothness of positioning results in the time domain.

[0043] Step 4: Multidimensional Measurement Covariance Weighting Based on the four factors in step 3, the diagonal elements of the measurement covariance are updated using a unified weighted formula. : (18) in, Let V be the set measurement variance. Through a product-like form, this formula can simultaneously respond to the coupled effects of multiple error sources. For example, the variance of a newly emerging base station that is both low-elevation and generates significant innovation will be greatly inflated, thereby reducing its impact on the filter.

[0044] Step 5: Kalman Measurement Update Using weighted measurement covariance Calculate Kalman gain : (19) in, This is the filtered observation matrix; Correcting the state vector using Kalman gain and posterior covariance matrix : (20) (twenty one) in, Represents the identity matrix.

[0045] Final output The position components in the equation are used as the high-precision positioning result for the current epoch.

[0046] Step 6: Epoch Iteration and Termination Determination Determine whether the current localization task has ended or whether there is still observation data input for subsequent epochs. If there is still data, then set the time epoch. And estimate the posterior state at the current time. and posterior covariance matrix As prior information for the next epoch, return to step 2 for the next round of recursive calculation; if there is no data, end the current positioning.

[0047] This invention conducted experiments to verify the algorithm. Considering that single satellite navigation or 5G positioning is difficult to use as a high-precision benchmark, this experiment selected Adaptive Kalman Filter (AKF) as the comparison benchmark. The positioning error calculated by AKF was compared with the positioning error calculated by AKF combined with Multi-Dimensional Measurement Covariance Shaping (MD-MCS) to verify the improvement of the invention in terms of accuracy and robustness.

[0048] In the experiment, this embodiment constructed a highly realistic urban canyon simulation scene, such as... Figure 3 As shown in Figure 4, the scenario includes seven high-rise buildings ranging in height from 110 to 160 meters, forming a typical deep canyon road. The simulation shows vehicles traveling along a closed trajectory approximately 3.2 kilometers long, with a simulation duration of 423 seconds. Regarding signal source configuration, a 5G network consisting of 15 base stations and a constellation of 13 BeiDou satellites using a hierarchical distribution strategy were deployed, with a minimum visible elevation angle set at 15°. The receiver clock bias is modeled as a first-order autoregressive process with an autoregressive coefficient of 0.995 and a process noise of 0.5m. The BeiDou pseudorange error includes 0.9m of thermal noise and 0.4m of flicker noise. The 5G parameter settings adopt a hybrid error model that conforms to the 3GPP standard. The 5G observation error is set to 18ns under line-of-sight conditions, with an equivalent distance of approximately 5.4m. Under non-line-of-sight conditions, a three-mode Gaussian mixture model is adopted, with the weights of the three components being 0.55, 0.30, and 0.15, respectively. The corresponding mean deviations are 15m, 50m, and 120m, respectively, and the time correlation coefficient is set to 0.95.

[0049] The specific parameter configuration of the MD-MCS module used in the experiment is as follows: Outlier detection threshold based on MAD Set to 2.6, weighted index Set to 2.6; 5G horizontal geometric reference value Set to 4.0, smooth weights Set to 0.2; Vertical sensitivity reference value Set to 0.7, vertical expansion index The initial expansion factor is set to 1.4; for newly emerging links, an initial expansion factor is set. , Smoothing factors are 24.0 and 9.0. It is 0.25.

[0050] The specific error indicators for this experiment are as follows:

[0051] The experimental results in the table above show that, compared to the positioning solution obtained by the standard AKF method, the fused positioning data obtained after correction by the method of this invention has higher accuracy. The root mean square errors of the three axes (east, north, and celestial) are reduced by 15%, 20.5%, and 6.8%, respectively. At the same time, the positioning error distribution is more convergent. Taking the horizontal direction as an example, the 50th percentile error in the east direction decreases from 0.70m to 0.55m, and the 90th percentile error decreases from 1.88m to 1.65m. The 50th percentile error and 90th percentile error in the north direction decrease from 0.59m and 1.73m to 0.45m and 1.43m, respectively. This indicates that the method of this invention not only improves the average accuracy but also more effectively suppresses tail errors, enhancing the robustness of the system in complex environments.

[0052] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A Kalman filter based Beidou+5G multi-dimensional measurement covariance weighted positioning method, characterized in that, include: Step 1: Defining the state vector for the Kalman filter Step 2: Defining the measurement vector for the Kalman filter Step 2: In the current epoch The pseudorange observation value based on Beidou satellites The distance observation value equivalent to the arrival time of 5G base stations is Observation equations for BeiDou satellite and 5G measurement values ​​were constructed respectively; Using the state estimation of the previous epoch Covariance Matrix Combined with the state transition matrix Update to obtain the current epoch. Next, the predicted state vector and predicted covariance : (4) (5) in The noise covariance matrix of the previous time step; Construct a nonlinear observation function composed of the BeiDou pseudorange observation equation and the 5G TOA observation equation. Assume there are a total of at time k Beidou satellites and One 5G base station, The expression takes the following form: (6) in, ... The first one... the second one The distance between a BeiDou satellite and the predicted location of the user terminal. , These are the ionospheric and tropospheric delays of the first Beidou satellite, respectively. , The first The ionospheric and tropospheric delay of a BeiDou satellite; , The first and the second, respectively. The distance between a 5G base station and the predicted location of a user terminal; Represents the speed of light; express For receiver clock bias; According to formula (6), each state variable in the matrix is ​​replaced with the predicted state variable at the current time k to obtain the observed predicted value at the current epoch. ; Simultaneously, the actual collected BeiDou pseudorange and 5G TOA data are combined into a measurement vector. Calculate the innovation vector The formula is as follows: (7) Step 3: Calculate the covariance inflation factor, including the outlier factor for each observation i. Horizontal geometric factors Vertical sensitivity factor and link continuity factor ; Step 4: Based on the covariance inflation factor in Step 3, update the diagonal elements of the measurement covariance using the unified weighting formula. : (18) in, The set measurement variance; Step 5: Kalman Measurement Update Using weighted measurement covariance Calculate Kalman gain : (19) in, This is the filtered observation matrix; Correcting the state vector using Kalman gain and posterior covariance matrix : (20) (21) in, Represents the identity matrix; Final output The positional components in the equation are used as the high-precision positioning result for the current epoch. Step 6: Epoch Iteration and Termination Determination Determine whether the current positioning task has ended or whether there is still observation data input for subsequent epochs: If there is still data, then set the time epoch. And estimate the posterior state at the current time. and posterior covariance matrix As prior information for the next epoch, return to step 2 for the next round of recursive calculation; if there is no data, end the current positioning.

2. The BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering as described in claim 1, characterized in that, Outlier inflation factor The calculation process is as follows: The dispersion of the innovation is assessed using the median absolute deviation statistic; first, the standardization factor is calculated. : (8) in, This indicates taking the median of the data; Calculate normalized new information Outlier inflation factor Defined as: (9) in, Indicates an exponent; Represents the information vector of the i-th observation; This indicates the set threshold.

3. The BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering as described in claim 1, characterized in that, Horizontal geometric factor The calculation process for the horizontal geometric distribution quality of 5G observations is as follows: Calculate the horizontal precision factor corresponding to the 5G observation matrix The calculation formula is as follows: in, and The weight coefficient matrix of the 5G observation equation is shown below. The eastward and northward variance components on the diagonal. Kalman filter observation matrix The submatrix corresponding to 5G observations; Define the original scaling value : (10) in, This represents the horizontal geometric reference value for 5G base stations; To avoid drastic fluctuations in this factor, an exponential moving average is used. Smoothing process is performed to obtain : (11) in, Indicates smoothing weights; Horizontal geometric factor Defined as: (12)。 4. The BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering as described in claim 1, characterized in that, Vertical sensitivity factor For 5G observation, the calculation process includes: First, extract the current 5G observations into the Kalman filtered observation matrix. Submatrix corresponding to 5G observations The third element of the corresponding row is denoted as the vertical component of the gaze vector. Calculate its absolute value Meanwhile, define vertical geometric indices. : (13) in, This represents the vertical reference value for 5G base stations; Define the original vertical penalty value for: (14) in, Indicates the vertical expansion index; Vertical sensitivity factor Defined as: (15)。 5. The BeiDou + 5G multidimensional measurement covariance weighted positioning method based on Kalman filtering as described in claim 1, characterized in that, Link continuity factor The calculation process is as follows: For observation i, define a smoothing variable for the current epoch k. For BeiDou satellites, if it is a newly emerging satellite link, then ,otherwise For 5G base stations, a recursive smooth update is adopted as follows: (16) in, As a new link smoothing factor, , This is the initial expansion coefficient; if it is a newly emerging 5G base station link, then... ,otherwise ; The link continuity factor is defined as: (17)。

Images