Improving the Stochastic Model for VRS Network-Based GNSS Surveying

Abstract The VRS network-based technique has become the main precise GNSS surveying method especially for medium-range baselines (approximately 20-70 km). The key concept of this approach is to use the observables of multiple reference stations to generate the network correction in the form of a virtual reference station for mitigating distance-dependent errors including atmospheric effects and orbital uncertainty at the user’s location. Numerous GNSS data processing strategies have been adopted in the functional model in order to improve both the positioning accuracy and the success of ambiguity resolution. However, it is impossible to completely model the aforementioned errors. As a result, the unmodelled residuals still remain in the virtual reference station observables when the least squares estimation is employed. An alternative approach to deal with these residuals is to construct a more realistic stochastic model whereby the variance-covariance matrix is assumed to be homoscedastic. This research aims to investigate a suitable stochastic model used for the VRS technique. The rigorous statistical method, MINQUE has been applied to estimate the variance-covariance matrix of the double-difference observables for a virtual reference station to rover baseline determination. The findings of the comparison to the equal-weight model and the satellite elevation-based model indicated that the MINQUE procedure could enhance the positioning accuracy. In addition, the reliability of ambiguity resolution is also improved.