The Semiparametric Regression Model In The Application Of Clock Bias Modeling Research

Precise Point Positioning is a technology that uses precise satellite ephemeris and satellite clock bias(SCB) from the IGS ground tracking stations, thus according to the carrier phase and the pseudorange observations collected by a single GPS receiver to calculate the exact coordinates. The timeliness of precise satellite clock bias is one of the key issues to realize the applications of real-time and high precision of precise point positioning(PPP). However, the application of real-time PPP technology has been restricted without the real-time and extrapolation satellite clock bias provided by IGS. Therefore, the application research of high-precision GPS satellite clock error prediction plays an outstanding role in realizing the real-time precise point positioning.Firstly, the time-frequency characteristics of satellite atomic clocks are analyzed, secondly, kinds of common prediction models are realized and method of risk appraising and evaluation, finally, the pros and cons of each model are summarized according to specific examples. On this basis, the semiparametric regression model is used in modeling and predicting, the main research contents of this paper as follows:(1)Analysis and research the time-frequency characteristics of satellite atomic clocks. The kinds of evaluation index which include the accuracy, stability and drift of frequency characteristic of the satellite atomic clocks are introduced. The Hadamard Variance and Modified Allan Variance are used to analyze and evaluate the stability of satellite atomic clocks selected to be modeling.(2)The characteristics of commonly used clock error prediction models are summarized.The several commonly used prediction algorithms are realized with the aid of precise clock file download form IGS, the pros and cons of each models are summarized according to specific examples. On the basis of this, the methods of classic weight and effective weight are used to build the integrated model.(3)The theory and calculation of semi-parametric regression model are discussed. The regular matrix R and smoothing parameter are determined by writing a MATLAB program. Combined the simulated instance, the penalized least squares estimation method based on the regular matrix and natural spline function are realized to summary and analysis the calculation characteristics of semi-parametric regression model.(4)The algorithm of the wavelet assisted quasi-accurate detection and penalized least square estimation is proposed to calculate the pollution semi-parametric model in the case which abnormal components, systems, and stochastic components coexist in the same observation samples; For the outliers in the satellite clock bias,the algorithm of wavelet assisted quasi-accurate detection which can posit and estimate the gross error of semi-parametric model.(5)The application of semi-parametric model in satellite clock error is mainly studied in modeling and predicting. The semi-parametric model can divide satellite clock bias into parameters component and nonparametric component by means of mathematical methods. Then, the polynomial model and the fourier trigonometric function model are built respectively according to the characteristic of the two components. Through the two models matched two data sources that can reduce the influence of model error, in the end the accurate modeling and forecasting results are obtained.