Research And Implementation Of Precise GPS Positioning Data Processing Algorithm Based On Refined Stochastic Model

The precision of GPS positioning is dependent on the functional and stochastic models,which carrier phase measurements are often used to achieving high accuracy results.Simplified equal-weighting stochastic model is now widely used in the data process of highprecision GPS positioning. Contaminated by different error sources, GPS measurements fromdifferent satellites or epochs have different precision and are spatial-temporally correlated.The simplified stochastic equal-weighting model implies that all the measurements are withthe same precision which deviates from reality. It is important to improve the data processingalgorithm for precise GPS positioning by defining a realistic stochastic model, which canpromote the precision and reliability of the baseline solution, especially under complicatedcircumstances. This research focuses on the algorithm of precise GPS positioning dataprocessing based on the definition of stochastic model. Major contents and conclusions are asfollows:1．A comparison of carrier phase quality indicators such as elevation, signal-to-noise andleast square residuals has been carried out. It appears that the three quality indicators of GPSmeasurements change in the same trend and affect the precision of the baseline solution. Thesine and exponential stochastic models are introduced after analyzing the relationshipbetween satellite elevation and standard deviation of measurements. The signal-to-noiseinformation is successfully extracted from GPS observation files with TEQC, which is usedto construct a SNR stochastic model. Experiment results show that the satellite elevation andSNR stochastic models both provide a significant improvement over the equal-weightingstochastic model.2．A BIQUE procedure for the estimation of GPS stochastic model has been presentedbased on the principles of best invariant quadratic unbiased estimation of variance-covariancecomponents. The design and implementation of the BIQUE module have been carried out.The reliability of the ambiguity are significantly improved after BIQUE procedure during thefloat solution. It is necessary to restart the BIQUE procedure to achieve high precision fixedsolution. Several experiments have been implemented. For static and rapid static relativeGPS positioning, both the reliability of ambiguity and the precision of baseline aresignificantly improved.3．A RBIQUE procedure for the estimation of GPS stochastic model has been proposedbased on the theory of robust parameter estimation for correlated observations. The designand implementation of the RBIQUE module have been carried out. It is demonstrated bynumerical example that the RBIQUE procedure for GPS stochastic model is feasible and caneffectively control the influence of outliers to estimation of variance-covariance components.The reliability and the precision of baseline solution are both further improved, especially thevertical component.