A study on predicting network corrections in PPP-RTK processing

Abstract

In PPP-RTK processing, the network corrections including the satellite clocks, the satellite phase biases and the ionospheric delays are provided to the users to enable fast single-receiver integer ambiguity resolution. To solve the rank deficiencies in the undifferenced observation equations, the estimable parameters are formed to generate full-rank design matrix. In this contribution, we firstly discuss the interpretation of the estimable parameters without and with a dynamic satellite clock model incorporated in a Kalman filter during the network processing. The functionality of the dynamic satellite clock model is tested in the PPP-RTK processing.Due to the latency generated by the network processing and data transfer, the network corrections are delayed for the real-time user processing. To bridge the latencies, we discuss and compare two prediction approaches making use of the network corrections without and with the dynamic satellite clock model, respectively. The first prediction approach is based on the polynomial fitting of the estimated network parameters, while the second approach directly follows the dynamic model in the Kalman filter of the network processing and utilises the satellite clock drifts estimated in the network processing. Using 1. Hz data from two networks in Australia, the influences of the two prediction approaches on the user positioning results are analysed and compared for latencies ranging from 3 to 10. s. The accuracy of the positioning results decreases with the increasing latency of the network products. For a latency of 3. s, the RMS of the horizontal and the vertical coordinates (with respect to the ground truth) do not show large differences applying both prediction approaches. For a latency of 10. s, the prediction approach making use of the satellite clock model has generated slightly better positioning results with the differences of the RMS at mm-level. Further advantages and disadvantages of both prediction approaches are also discussed in this contribution.