Research On Key Technologies Of Multi-GNSS Precise Data Processing

GNSS precise data processing is closely related to the application of GNSS in the field of scientific research and engineering.After years of research,the theory and algorithms of GNSS precise data processing have basically been matured and are booming towards higher precision and more diverse applications.However,with the development and maturity of GNSS systems,the number of global or regional reference stations has increased dramatically and real-time data applications have emerged.GNSS precise data processing faces new opportunities and challenges.Firstly,for large-scale reference station positioning,the scale of calculation is huge,and the calculation is time-consuming.The limited hardware upgrade space cannot meet the urgent needs for calculate ability and computing resources.Secondly,the calculation of satellite-related parameter calculations in precise data processing increases,and the calculation time becomes longer.Because GPS and GLONASS systems are increasingly improved,and emerging navigation systems such as BDS and Galileo are gradually developing,more navigation satellites participate in global application services.Finally,real-time or quasi-real-time research and applications are increasingly demanding low latency and concurrency.With the development of computer hardware and parallel data processing technology,new solutions have been introduced for GNSS large-scale,multi-system,low-latency high-precision data processing.Based on this,this paper studies the key technologies of multi-system GNSS parallel precise data processing.TheMPI,OpenMP and Pthread parallel models are combined with GNSS precise orbit determination,satellite clock offset estimation,large-scale precise positioning and atmospheric parameter modeling.It is designed and implemented based on the cluster platform.The main research contents include:1)Aiming at the problem of singularity in satellite orbit theory,In-depth analysis and research on the Lagrange/Gauss satellite equations of motion without singular points,starting from the original Lagrange/Gauss motion equation and their physical significance,considering the three situations of circular orbit,circular equatorial orbit and equatorial orbit,a new Lagrange/Gauss satellite motion equation without singular point is derived,and the continuity of the equation is discussed.This equation completely eliminates the zero factor and fundamentally solves the singularity problem of the satellite motion equation.2)Aiming at the multi-system,high-time-efficiency calculation requirements for calculating GNSS satellite orbit and clock offset,this paper proposes a multi-system GNSS precise orbit and clock parallel determination method based on MPI/OpenMP parallel programming technology.At first,this paper researches on different parallel modes and methods for the main time-consuming processing based on the basic principles of GNSS precise orbit and clock offset calculation.The influence of the number of threads and the number of processes on the computational efficiency,as well as the case of single-system GNSS/multi-system GNSS and more stations,the applicability and effectiveness of parallel algorithms have also been analyzed.Experimental results show that the use of multi-process and multi-thread technology can improve the efficiency of GNSS precise orbit determination and clock offset estimation.The hybrid parallel method can increase the efficiency most,and the more threads and processes,the greater the acceleration ratio is.The parallel method has more advantages in improving efficiency of calculating the orbit and clock offset with multiple stations and multi-systems.The calculation efficiency of the four-system orbit determination is increased by about 30%,and the efficiency of the four-system clock offset estimation is increased by about 59%.And the accuracy loss of both can be ignored.On the base of that,aiming at the problem of the accuracy loss of GNSS real-time orbit with extrapolation time,this paper proposes a new method of multi-period hybrid parallel real-time orbit fast determination.The basic principle and processing flow of MPI multi-periodnormalequation accumulation and OpenMP multi-thread parameter elimination are deduced and designed.Finally,the parallel method of multi-system GNSS real-time orbit determination is realized.The validity of the method is verified by measured data.This method can increase the orbit update frequency from 6 hours to 1 hour.The precision comparison and analysis results of SDU ultra-fast orbit show that the accuracy of real-time orbit part of 1h result is 30%higher than that of 6h result.Compared with similar international orbits,the accuracy and stability of the real-time part of the SDU ultrafast orbit are better.The RMS of GPS,GLONASS,Galileo,and BDS orbit are 3.21 cm,5.08 cm,5.56 cm,and 11.83 cm,respectively,which have the same accuracy as international products.3)Aiming at the problems of too many parameters to be estimated and low calculation efficiency in the precise double-difference network positioning,this paper proposes a new method of equivalent parallel calculation combining MPI technology and equivalence theory.Thereservation of covariance matrix,elimination of unknowns and parallelism of the equivalence algorithm are derived.The efficiency and precision of the proposed method is verified in experiments with double-differenced observation equations.The algorithm can selectively reduce unknown numbers with negligible precision loss(approximately 10-9 metre)in applications.The efficiency of processing 100 IGS stations is improved significantly while the precision loss can also be ignored with the parallel equivalence algorithm.The improvement of computational efficiency can reach 56%.Further comparison of efficiency shows that the proposed approach has higher computational efficiency compared with the serial least squares and Gauss-Jordan parallel method and this advantage is more obvious as computational scale increases.Furtherexperimentson cluster with multi-node show that the more nodes used,the higher efficiency it can achieve.4)Aiming at the problems of huge calculation amount in large-scale post positioning and strong concurrency in large-scale real-time processing,this paper uses different parallel computing ideas to design and implement GNSS large-scale post and real-time parallel processing methods.On the one hand,the MPI/OpenMP hybrid parallel method is used for the large-scale GNSS post positioning.The basic principle of the method is derived and analyzed in detail,and the multi-process and multi-thread processing is designed and implemented.Through the measured data of 270 stations,the performance of the algorithm on efficiency improvement is verified,and the efficiency improvement is up to 53.6%.On the other hand,Pthread multi-thread parallel method is used for large-scale GNSS real-time positioning.The characteristics of concurrent multiple data streams and real-time positioning processing are analyzed.On this basis,the process and experimental scheme of Pthread multi-thread parallel real-time positioning are designed.Finally,with the help of MPI and Perl,the real-time multi-threaded parallel positioning of 1500 reference stations was realized based on the real-time data stream.5)In order to solve the problem of huge calculation amount and long time-consuming in the construction of a multi-source meteorological data troposphere model with high spatial and temporal resolution,this paper designs a parallel tropospheric calculation method,based on which a 1°*1° tropospheric grid model of China considering day period information is realized.The early tropospheric model based on multi-source meteorological data was limited to the spatial and temporal resolution of the data,and the modeling based on the data of the day resolution only considered the annual and semi-annual cycles.Based on the 1-hour time resolution tropospheric parameters extracted by the latest data ERA5,this paper proposes a model that takes into account changes in the troposphere within a day.Based on the MPI parallel programming technology,the model is constructed on the cluster platform.IGS troposphere products and ERA5 troposphere were selected to evaluate the accuracy of the model.The results show that the model has higher accuracy in northern China than in southern China.Compared with GPT2w model and low time resolution modeling results,the new model has higher accuracy.This proves that parallel computing has a positive significance for tropospheremodeling with high spatial-temporal resolution to improve accuracy.In addition,the accuracy and spatial-temporal distribution characteristics of the tropospheric troposphere extraction in China are analyzed,which provides a reference for the calculation of high-precision troposphere and related parameters in China.6)Aiming at the improvement of global ionospheric model building efficiency,a parallel method of GNSS ionosphere modeling was realized.The increase of global IGS reference stations and the development of GNSS systems have increased the amount of calculation for the GNSS ionospheric modeling and reduced the efficiency of data processing.Therefore,this paper implements a parallel extraction and modeling of GNSS ionosphere.It is found that the time-consuming process of building the ionospheric model is the extraction of the TEC and the model solution.Based on this,the parallel processing method is used to optimize the entire process.The computational efficiency of using different parallel strategies and hybrid parallel methods in different processes is further analyzed.Experimental results show that parallel computing can significantly improve the efficiency of GNSS ionospheric model construction.Compared with serial schemes,multi-process schemes,multi-thread schemes,and hybrid parallel schemes,hybrid parallel computing strategies have the highest efficiency.Compared with serial schemes,multi-process schemes,multi-thread schemes,and hybrid parallel schemes,hybrid parallel computing strategies have the highest efficiency.It provides a new strategy for parallel modeling of GNSS ionosphere,which is beneficial to the performance improvement of ionosphere products.