Application Of Multi-Type Observations Fusion Modeling In Low-Cost Positioning Cycle Slips Detection

Carrier phase observation is an important data source for GNSS(Global Navigation Satellite System,GNSS)to achieve high precision positioning.It is important to ensure its correctness and completeness.So,we need to do quality analysis and outlier detection and other pretreatment.As outliers in carrier phase observations,the detection methods of single-frequency and dual-frequency data are different.In this thesis,the existing single-frequency and dual-frequency detection methods are improved and verified by adding different types of data and building different models.Specific work is as follows:(1)The main observation quantity and error source of GNSS are expounded.The quality analysis of GNSS observation data collected by receivers of different levels is carried out in detail,which mainly includes three aspects: multipath effect,signal-tonoise ratio and ionospheric delay.The anti-multipath capability and signal quality of receivers of different levels are analyzed,and the relationship among multipath,height Angle and signal-to-noise ratio is expounded.(2)As for the existing single-frequency detection methods,Doppler observations are fully utilized.According to the idea of sliding window,the L-order polynomial model of pseudo-range variation is obtained by using n Doppler samples in the window,namely Dn Pl model.Instead of the traditional trapezoidal integral method,the carrier phase can be predicted at any time(non-observed epoch).It can assist the cycle slip detection of carrier phase.So we can detect cycle slip as small as one cycle(3)Aiming at the problem that the original D5P2 model cannot detect minor cycle slip for low sampling rate data,a Time Differenced Carrier Phase(Time Differenced Carrier Phase,TDCP)co-modeling method is proposed to solve the parameters.The model is improved to Cm Dn Pl model.Cm refers to the carrier phase observation at m epochs.Through theoretical and experimental analysis of the model,it is found that the precision of cycle slip detection value of the C5D5P2 model is about 80% higher than that of D5P2,which can detect the cycle slip of low sampling rate data from small to one cycle.(4)Aiming at the problem that the accuracy of the traditional phase pseudo-range reduction method is limited by the accuracy of the pseudo-range,a method of Doppler smoothing pseudo-range /phase combination is proposed.The pseudo-range smoothed based on Hatch filter is used to replace the original pseudo-range for phase combination to detect cycle slip.Experiments show that the accuracy of this method is improved by about 90% compared with the traditional phase pseudo-range reduction method,and it is of great significance for the cycle slip detection of low-cost receivers.(5)Aiming at the problem that the traditional non-geometric distance phase combination(Geometry-free,GF)can not detect the sensitive cycle slip and can not separate the cycle slip of each frequency point.By combining the proposed Doppler smoothing pseudo-range/phase group method with GF method,the cycle slip is calculated by the two methods simultaneously.The experimental results show that this method can detect the sensitive cycle slip combination successfully.And the different frequency cycle slips are separated successfully.It makes up for the shortcomings of the traditional GF method.(6)Aiming at the problem that the cycle slip of each frequency point cannot be separated by the two-frequency code phase grouping method(Melbourne-Wübbena,MW),Kalman filter is used to filter the observed value of the pseudo-range narrowlane combination,thus improving the accuracy of MW detection value,so that the MW detection value can be directly used as the estimate of cycle slip,without the subsequent recursive process.And the cycle hop of different frequency points can be separated successfully.(7)Taking advantage of the sparsity of cycle slips,a cycle slip detection method based on sparse regularization is proposed,and all epoch cycle slip estimates are simultaneously taken as parameters to be estimated for iterative solution.L1 regularization was used to ensure the sparsivity of the estimated parameters,and generalized cross validation(GCV)criterion was used to determine the hyperparameters of the cost function.Finally,the fast iterative shrinkage threshold algorithm(FISTA)was used to solve the estimated parameters.Experimental verification shows that this method has good flexibility.For the case of frequent cycle slip(15% cycle slip),the success rate of added cycle slip detection can reach nearly100% by properly adjusting the super parameter to achieve a balance between the leakage rate and the false detection rate.This is a meaningful attempt.