Research On Method Of GNSS Observation Precision Estimation And Stochastic Model Refinement

Since the development of GPS in the United States in 1973,the former Soviet Union,China and the European Union have launched the construction of GLONASS,BDS and Galileo to meet the needs of users for high-precision navigation,positioning and timing etc.Nowadays,satellite navigation and positioning technology has been widely used in various fields such as national security,economy,society and lives.Accurate and reliable solution can be obtained only when both the functional model and the stochastic model are accurately constructed in GNSS data processing.Functional model reflects the functional relationship between pseudorange and carrier phase observations and unknown parameters such as coordinate parameters,atmospheric delay and ambiguity parameters.While stochastic model focuses on describing the prior statistical characteristics of observations.At present,the research on GNSS data processing function model for multi-frequency and multi-system is relatively mature,while the study and application of stochastic model is comparatively rare.Research on precision estimation method of multi-mode and multi-frequency GNSS observations can provide necessary priori information for satellite navigation and positioning.More over,to develop objective and practical stochastic model and establish reasonable weight ratio of observations can provide important guarantee for realization of high-precision positioning and its application.Therefore,this paper focuses on multi-system observation precision estimation and refined method of stochastic model,and analyzes the application effects of different stochastic models based on deformation monitoring data of dams and super-large bridges.The specific contents are as follows:(1)A reliable estimation method for carrier phase single-difference(SD)residuals and data processing strategy of multi-mode GNSS are proposed.Aiming at the problem of ambiguity fixed for estimation of SD residual between stations,TurboEdit algorithm is used to traverse SD M-W/GF sequences,detect cycle slips in each arc segment and smooth the original sequence.The results show that: the sequence smoothing avoids about 2 cycles original ambiguity deviation and TurboEdit algorithm can improve the accuracy in fixing ambiguity with the combination of M-W and GF.Then,the mean drift strategy is used to detect and eliminate gross errors in OMC sequence.Finally,the reliable carrier phase SD residual sequence can be obtained based on clean OMC sequence with least square estimation method.(2)The variation rule of SD residuals of multi-mode and multi-frequency GNSS observations is analyzed.Based on measured data of IGS CUT0 station,we analyze the variation of GNSS observations SD residuals of each mode and each frequency.The results show that the variation trend of SD residuals in different frequencies and different systems is not uniform.The L2 SD sequence of each system is more sensitive to elevation variation and the signal quality is worse than that of L1.Compare zero baseline experimental results of same receivers and hybrid receivers,it is found that the signal noise is amplified in various degrees due to the difference of internal configuration and decoding mode in receivers.The SD sequence of GLONASS system does not converge due to the influence of inter-frequency deviation.(3)The refined method of stochastic model based on elevation and carrier-to-noise ratio(CNR)is studied and the refined models are build.Furthermore,the application effect of refined stochastic model based on deformation monitoring data of dams and extra-large bridges is evaluated.Based on above estimated SD residual sequence,the precision of dual-frequency observations in different systems is estimated,and different mathematical functions are used to express the actual variation characteristics of pseudorange and carrier phase observation precision with elevation and normalized CNR.Based on measured data of Xilongchi Reservoir in Shanxi Province and Forth Road Bridge in England,we verified the application effects of different empirical and refined stochastic models.The results show that the effects of empirical and refined stochastic models in dynamic and static mode is equivalent when in good observation environment.While in the case of poor observation environment,refined stochastic models can enormously reduce external environment effect on dam long-term deformation monitoring,and coordinate components residual in the direction of occlusion change significantly.For real-time dynamic monitoring in super-large bridge,low-frequency noise of deformation sequence in occluded station can be significantly reduced with refined CNR model,which has great practical application value.