Role of the RLS-VCE-estimated stochastic model for improvement of accuracy and convergence time in multi-GNSS precise point positioning

Abstract

• We solve the PPP problem using RLS. • The VCE method used to estimate the appropriate stochastic model. • The use of the RLS-VCE method for stochastic model estimation has a significant impact on PPP performance. • The improvement of the PPP performance is investigated by increasing accuracy and reducing convergence time. • The combination of multi-GNSS observations has an impact on PPP positioning performance. We present a method for PPP adjustment based on Recursive Least Square Estimation (RLSE). This method estimates the stochastic model of GPS, GLONASS, Galileo and BeiDou observations by using RLS-Variance Components Estimation (RLS-VCE). The accuracy and the convergence time of horizontal, altitudinal and three-dimensional components is compared by using both nominal and RLS-VCE method so that the estimated accuracy for these coordinate components in suggested method is more accurate than nominal method. The accuracy of horizontal, altitudinal and three-dimensional components in this method are 1.3, 1.25 and 1.8 cm respectively in comparison to nominal method with 4, 3.8 and 5.6 cm. By considering 15 cm as the convergence condition, convergence time is estimated 3 and 5.3 min for suggested and nominal methods respectively. The obtained RMSE after convergence time in proposed and nominal methods are 7.1 and 9.2 cm respectively. According to the presented results, when Multi-GNSS observations are combined, the lowest RMSE are estimated for the coordinate components. As time passes, the stochastic model becomes more stable, therefore, the presented method will be more accurate.