Theory And Methodology Of Multi-GNSS Undifferenced And Uncombined Precise Point Positioning

As a global positioning technology developed in the late 1990 s,precise point positioning(PPP)can provide high-precision absolute position on a global scale in a flexiable and efficient way.Meanwhile,the information such as receiver clock offset and the atmospheric delays etc is retained in the PPP processing.Therefore,PPP has been widely applied in precise positioning,timing,atmospheric sensing,early warning of earthquakes and tsunami and other related Earth science research.In general,the traditional PPP model is based on the ionospheric-free observations,which is through the specifically linear combination of the dual-frequency GNSS observations to eliminate the first-order ionospheric delays.While,PPP with the undifferenced and uncombined models has many advantages over the ionospheric-free PPP,which directly uses the original GNSS observations and is more flexiable and suitable for the current and future multi-constellation and multi-frequency GNSS PPP processing.Many studies on the theory and applications of GPS undifferenced and uncombined PPP have been carried out,however,studies on multi-GNSS undifferenced and uncombined PPP are quite limited.Moreover,multi-GNSS undifferenced and uncombined PPP will bring more challenges,such as GNSS biases,the handling schemes of ionospheric delay parameters etc.In addition,with the spatial-temporal characteristics of the ionosphere,the research on positioning performance of introducing external ionospheric delay constraints on the ionospheric delay paramters is still very limited.On the other hand,to satify the increasing demand of real-time high-precision positioning and related applications,real-time PPP is another hot research topic nowadays.However,real-time PPP is mainly limited to simulated real-time mode or using the real-time products from a single analysis center.Therefore,it is necessary to study the algorithms and applications of multi-GNSS PPP with undifferenced and uncombined observations based on real-time streams.This dissertation focuses on the research topics such as “multi-GNSS”,“undifferenced and uncombined PPP”,“GNSS biases”,“real-time PPP” and carries out studies on the models,algorithms and applications.The main work and contributions of this thesis are listed as follows:(1)Based on full consideration of various GNSS biases,we establish a general multi-GNSS undifferenced and uncombined PPP model and develop the corresponding PPP processing software,which satisfies different signal modulation technology(e.g.,CDMA and FDMA)and different frequency(i.e.,single-,dual-,and multi-frequency),then design experiments to validate the accuracy and reliability of the model and software.(2)Considering that the conventional estimation scheme of GLONASS pseudorange inter-frequency bias(IFB)is neglecting IFB or based on the observations from GNSS tracking network in post-processing mode,this study proposes a method of simultaneous estimation of GLONASS IFB during PPP processing for single station.In this contribution,the impact of GLONASS pseudorange IFB on convergence performance and positioning accuracy of GLONASS-only and GPS + GLONASS PPP based on undifferenced and uncombined observation models is investigated.Through a re-parameterization process,the following four pseudorange IFB handling schemes were proposed: neglecting IFB,modeling IFB as a linear or quadratic polynomial function of frequency number,and estimating IFB for each GLONASS satellite.One week of GNSS observation data from 132 International GNSS Service(IGS)stations was selected to investigate the contribution of simultaneous estimation of GLONASS pseudorange IFB on GLONASS-only and combined GPS + GLONASS PPP in both static and kinematic modes.The results show that considering IFB can speed up the convergence of PPP using GLONASS observations by more than 20%.Apart from GLONASS-only kinematic PPP,the positioning accuracy of GLONASS-only and GPS + GLONASS PPP is comparable among the four schemes.Overall,the scheme of estimating IFB for each GLONASS satellite outperforms the other schemes in both convergence time reduction and positioning accuracy improvement,which indicates that the GLONASS IFB may not strictly obey a linear or quadratic function relationship with the frequency number.(3)The next focus is on proper modeling of the dynamics for inter-system bias(ISB)in multi-GNSS PPP processing.The current estimation strategies of ISB are mostly limited to the constant estimation.This paper proposed to adopt the random walk or white noise process to model the ISB parameters.First,the theoretical derivation demonstrates that the ISB stems from not only the receiver-dependent hardware delay differences among different GNSSs,but also the receiver-independent time differences caused by the different clock datum constraints among different GNSS satellite clock products.Afterwards,comprehensive evaluation of the influence of ISB stochastic modeling,i.e.,time constant,random walk process,and white noise process,on undifferenced and uncombined PPP performance,which was based on a one-month period(September 2017)Multi-GNSS Experiment(MGEX)precise orbit and clock products from four analysis centers(e.g.,CNES,CODE,GFZ,and WHU)and 160 MGEX tracking stations,was conducted.The results demonstrated that generally,the positioning performance of PPP in terms of convergence time and positioning accuracy with the final products from CNES,CODE,and WHU is comparable among the three ISB handling schemes.However,estimating ISB as random walk process or white noise process outperforms that as time constant when using the GFZ products.These results indicated that the traditional estimating ISB as time constant may be not entirely reasonable.To achieve more reliable positioning results,it is highly recommended to consider the ISB as random walk process or white noise process in the multi-GNSS PPP processing.(4)Based on a comprehensive consideration of various GNSS biases,such as GLONASS pseudorange IFB and GNSS ISB etc and the standard undifferenced and uncombined PPP models,the single-and dual-frequency ionospheric-unconstrained and-constrained(constraint from external ionospheric products)PPP models are deduced.Three commonly used strategies to constrain the virtual ionospheric observables are summed up,which are constant constraint,spatial-temporal constraint,and stepwise-relaxed constraint.First,the three strategies are selected to determine the suitable aprior variance of the virtual ionospheric observables.The experimental results demonstrated that the convergence performance of GPS single-frequency kinematic ionospheric-constrained PPP using the three strategies is obviously better than that of single-frequency ionospheric-unconstrained PPP.However,the positioning accuracy of PPP using constant constraint or spatial-temporal constraint after convergence is much lower than that of ionospheric-unconstrained PPP,while the positioning accuracy of PPP with stepwise-relaxed constraint is comparable to that of ionospheric-unconstrained PPP.The positioning convergence of multi-GNSS single-and dual-frequency kinematic ionospheric-unconstrained PPP and ionospheric-constrained PPP with stepwise-relaxed constraint is fully evaluated.The experimental results demonstrated that the external ionospheric constraints(such as GIMs)do not contribute much to dual-frequency PPP,while the convergence performance of single-frequency GPS-only and GPS + GLONASS PPP is significantly improved by more than 40%.Furthermore,the vehicle-borne kinematic experiment also demonstrated that adding external ionospheric constraints could significantly accelerate the convergence and improve the positioning accuracy of single-frequency PPP,while the performance improvement of dual-frequency PPP was very limited.(5)The current real-time PPP research and applications are mainly limited to simulated real-time mode or based on the real-time orbit and clock products from a single analysis center,this article aims at giving a comprehensive evaluation of IGS real-time service(RTS)orbit and clock products.The availability and quality of seven types of IGS RTS orbit and clock products in a one-month period(April 2017)was evaluated with respect to the ESA(European Space Agency)final products.In general,the availability of RTS products was above 95% for GPS and above 90% for GLONASS.Among these IGS RTS products,orbit comparisons showed that the 3D accuracy(root mean square,RMS)was about 3.2-7.9 cm for GPS satellites and 7.3-10.3 cm for GLONASS satellites.Afterwards,clock comparisons presented that the precision(standard deviation,STD)was 0.1-0.2 ns for GPS satellites and 0.2-0.4 ns for GLONASS satellites.Furthermore,based on the undifferenced and uncombined observation models,comprehensive evaluation of the impact of RTS products on real-time PPP processing,PPP tropospheric delay(PTD)estimation,and PPP ionospheric delay(PID)estimation was conducted.Comparative analysis demonstrated that the performance of precise positioning and PTD estimation using CNES real-time products generally outperformed that of the other real-time analysis centers.Furthermore,single-difference results of short baseline proved that PIDs were more precise than the traditional leveling ionospheric delays(LIDs).