Abstract:
In order to find an optimum stochastic modeling method for VRS (virtual reference station) systems, the principles of four stochastic models used in a real-time dynamic data processing system, including the standard stochastic model, the elevation-dependent model, the SNR (signal noise ratio) model and the self-adaptive model, were introduced, and their validity, advantages and disadvantages to a VRS system were analyzed from posterior variance, ADOP (ambiguity dilution of precision), F-ratio and filter residual. In addition, solutions to improve these models were given in the sight of statistics and their disadvantages. The results indicate that the filter residual and posterior variance obtained by the self-adaptive model change slightly around 0 and 1 respectively, showing the property of white noise. Furthermore, the advantage of the self-adaptive model in ADOP calculation is obvious, compared with the other models, however, it has an ordinary performance in F-ratio calculation. Totally speaking, the self-adaptive model should be the first candidate of a VRS stochastic model.