CN-121995410-A - Satellite estimation clock joint weight determining method based on mutual information and clustering

CN121995410ACN 121995410 ACN121995410 ACN 121995410ACN-121995410-A

Abstract

The invention relates to a satellite clock estimation joint weighting method based on mutual information and clustering, which comprises the steps of S1 reading original observation data of a monitoring station covering a service area, extracting altitude angle and signal-to-noise ratio parameters of each satellite and constructing the parameters into two-dimensional feature matrix data, S2 classifying the two-dimensional feature matrix data by adopting K-means clustering to obtain cluster distribution, S3 respectively calculating first mutual information between the altitude angle and the clusters by using a mutual information method to serve as first weight and second mutual information between the signal-to-noise ratio and the clusters to serve as second weight, S4 substituting the first weight and the second weight into a random model to form a joint weighting model, S5 using the joint weighting model to calculate double-frequency pseudo-range noise variance and carrier phase noise variance, S6 executing posterior quality control on real-time satellite clock difference Kalman filtering, carrying out Kalman filtering post-test residual error checking, finding out abnormal values and eliminating the abnormal values.

Inventors

LI HUI
ZHANG HAODONG
JIA CHUN
LI LIANG
QI BING

Assignees

哈尔滨工程大学

Dates

Publication Date: 20260508
Application Date: 20260211

Claims (8)

1. A satellite estimation clock joint weighting method based on mutual information and clustering comprises the following steps: Step S1), original observation data of a monitoring station covering a service area are read, and parameters of a height angle and a signal-to-noise ratio are extracted for each satellite and are constructed into two-dimensional feature matrix data; S2) classifying the constructed two-dimensional feature matrix data by adopting K-means clustering to obtain cluster distribution; Step S3) based on the obtained cluster distribution, respectively calculating first mutual information between the height angle and the cluster and second mutual information between the signal to noise ratio and the cluster by using a mutual information method, wherein the first mutual information is used as a first weight, and the second mutual information is used as a second weight; s4) substituting the obtained first weight and second weight into a random model, and establishing a refined joint weighting model; Step S5) calculating a double-frequency pseudo-range noise variance and a carrier phase noise variance by using a joint weighting model; Step S6) performing posterior quality control on the real-time satellite clock difference Kalman filtering, performing Kalman filtering posterior residual error detection by using the double-frequency pseudo-range noise variance and the double-frequency pseudo-range noise variance, finding out an abnormal value and rejecting the abnormal value.
2. The satellite estimation clock joint weighting method based on mutual information and clustering according to claim 1, wherein in step 1, a real-time clock difference estimation server is used for reading original observation data, and normalization processing are carried out on the acquired altitude angle and signal to noise ratio data of all frequency bands before a two-dimensional feature matrix is constructed.
3. The method for joint weighting and determining of satellite estimation clock based on mutual information and clustering according to claim 1, wherein the specific process of step 2 comprises the following steps: (1) Randomly initializing a cluster center: ; Where J denotes an objective function, x i denotes an i-th observation vector containing a height angle and a signal-to-noise ratio, μ j is a center of a J-th cluster, k denotes an optimal number of clusters, C j denotes a class in which a distance between data of an observation point and a cluster is closest, Representing the distance of the parameter; (2) Calculating the distance between the data of each observation point and the cluster center and finishing classification: (2) Wherein: (3) Wherein the method comprises the steps of Is the number of data points in the cluster, Is the sum of all data points within the cluster; (3) And (3) repeatedly executing the formulas (2) and (3), and terminating classification when the cluster center is not changed any more, so that data are distributed to corresponding clusters to obtain cluster distribution.
4. The method for jointly determining weights of satellite estimation clocks based on mutual information and clustering according to claim 1, wherein in the step 3, the method for respectively calculating the first mutual information between the altitude angle and the cluster and the second mutual information between the signal-to-noise ratio and the cluster by using the mutual information method is as follows: ; Wherein the method comprises the steps of For the first weight, MI (x, z) is the amount of mutual information between the altitude angle and the cluster, For the second weight, MI (y, z) is the mutual information quantity between the signal-to-noise ratio and the cluster, p (x, z) is the joint distribution probability of the altitude angle and the cluster, p (y, z) is the joint distribution probability of the signal-to-noise ratio and the cluster, p (x) is the edge probability distribution of the altitude angle, p (y) is the edge probability distribution of the signal-to-noise ratio, and p (z) is the edge probability distribution of the cluster.
5. The method for joint weighting and determining satellite estimation clock based on mutual information and clustering according to claim 4, wherein in step 4, the joint weighting model is: ; Wherein the method comprises the steps of In order to refine the noise variance after the refinement, Is the noise variance obtained using the altitude model, Is the noise variance obtained using the signal-to-noise model, For the first weight of the altitude angle in the joint weighting model, The second weight in the joint weighting model is the signal-to-noise ratio.
6. The method for joint weighting of satellite estimation clocks based on mutual information and clustering according to claim 5, wherein in step 5, the method for calculating the double-frequency pseudo-range noise variance and the carrier phase noise variance is as follows: ; Wherein the method comprises the steps of For the combined double-frequency pseudorange noise variance, For the combined carrier-phase noise variance, i represents the number of observations, Is the single-frequency noise variance of the pseudo range in the frequency band 1 calculated by the joint weighting model, Is the single-frequency noise variance of the pseudo range in the frequency band 2 calculated by the joint weighting model, Is the carrier phase single frequency noise variance at frequency band 1 calculated by the joint weighting model, Is the carrier phase single frequency noise variance at frequency band 2 calculated by the joint weighting model, For the wavelength of frequency 1 used by the satellite system, A wavelength of frequency 2 used for satellite systems.
7. The method for joint weighting of satellite estimation clocks based on mutual information and clustering according to claim 6, wherein in step 6, in the process of performing a posterior quality control, a posterior filtering measurement noise covariance matrix is first constructed as follows: ; where R is the measurement noise covariance matrix, For the combined double-frequency pseudorange noise variance, Is the combined carrier phase noise variance; secondly, introducing a measurement ratio to test a Kalman filtering post-test residual, and when the measurement ratio is larger than 9, recognizing the Kalman filtering post-test residual as an abnormal value, and eliminating the abnormal value, wherein the measurement ratio is as follows: ; where r m is the measurement ratio and V is the kalman filter post-test residue.
8. The method for joint weighting of satellite estimation clocks based on mutual information and clustering according to claim 7, wherein the checking of the Kalman filtering post-test residual errors is repeatedly performed until all Kalman filtering post-test residual error data passes the checking or the maximum number of iterations is reached.

Description

Satellite estimation clock joint weight determining method based on mutual information and clustering Technical Field The invention relates to the technical field of satellite navigation, in particular to a satellite clock estimation joint weight determining method based on mutual information and clustering. Background Along with the rapid development of the high-precision positioning and real-time service technology of the global satellite navigation system (Global Navigation SATELLITE SYSTEM, GNSS), the real-time clock error product is widely applied to scenes such as precise agriculture, intelligent transportation, unmanned aerial vehicle inspection, unmanned mine, automatic driving and the like. The scene not only requires centimeter-level or even higher positioning precision, but also provides higher requirements on the stability, reliability and robustness of a real-time calculation result. High-precision Real-time service represented by Real-time precision single point positioning (Real-TIME PRECISE Point Positioning, RT-PPP) technology requires continuously broadcasting correction products such as satellite clock error, orbit, deviation and the like to users in second-level time, and assists the users to quickly complete error correction and obtain stable high-precision positioning. However, in a complex observation environment with multiple GNSS fusion, due to inconsistent signal transmission characteristics of different systems and different frequency bands, the problems of unstable noise, increased low-quality observables and the like often occur in observation data under the actions of factors such as multipath interference, signal shielding, ionosphere turbulence and the like. If the random model cannot accurately reflect the observation quality difference, the posterior quality control is frequently triggered in the satellite clock error filtering process, the iteration times are greatly increased, the reliability of a real-time clock error product is further affected, and the real-time performance of the RT-PPP service is reduced. The current mainstream real-time clock error estimation method still generally adopts a traditional model based on an Elevation angle (Elevation angle), but the model only considers the geometric configuration advantage physically, and cannot describe the real measurement noise change caused by signal-to-noise ratio fluctuation, especially in a scene with increased abnormal observed quantity, weight distribution distortion is caused, so that the problems that the observed noise characteristic cannot be accurately reflected, the abnormal value is difficult to identify, the iteration number of real-time satellite clock error estimation is greatly increased and the real-time performance is insufficient caused by the abnormal observed quantity are caused. Disclosure of Invention In order to solve the above problems, the present invention provides a satellite estimation clock joint weighting method based on mutual information and clustering, comprising: Step S1), original observation data of a monitoring station covering a service area are read, and parameters of a height angle and a signal-to-noise ratio are extracted for each satellite and are constructed into two-dimensional feature matrix data; S2) classifying the constructed two-dimensional feature matrix data by adopting K-means clustering to obtain cluster distribution; Step S3) based on the obtained cluster distribution, respectively calculating first mutual information between the height angle and the cluster and second mutual information between the signal to noise ratio and the cluster by using a mutual information method, wherein the first mutual information is used as a first weight, and the second mutual information is used as a second weight; s4) substituting the obtained first weight and second weight into a random model, and establishing a refined joint weighting model; Step S5) calculating a double-frequency pseudo-range noise variance and a carrier phase noise variance by using a joint weighting model; Step S6) performing posterior quality control on the real-time satellite clock difference Kalman filtering, performing Kalman filtering posterior residual error detection by using the double-frequency pseudo-range noise variance and the double-frequency pseudo-range noise variance, finding out an abnormal value and rejecting the abnormal value. The invention has the beneficial effects that: The invention provides a satellite estimation clock joint weighting method based on mutual information and clustering, which utilizes unsupervised K-means clustering to construct a signal quality cluster, and quantifies the contribution degree of a satellite altitude angle and a signal-to-noise ratio to observed quantity quality based on a mutual information (Mutual Information, MI) method, thereby constructing a joint weighting model which accords with actual observation distribution characteristics, and using the mo