On accuracy and reliability of instantaneous network RTK as a function of network geometry, station separation, and data processing strategy

The network-based approach to kinematic GPS positioning significantly increases the distance, over which carrier-phase ambiguity resolution can be performed. This can be achieved either by introducing geometric conditions based on the fixed reference locations, and/or through the use of reference network data to estimate the corrections to GPS observations that can be broadcast to the users. The Multi Purpose GPS Processing Software (MPGPS) developed at The Ohio State University uses the multiple reference station approach for wide area and regional differential kinematic GPS positioning. The primary processing algorithm uses the weighted free-net (WFN) approach with the distance-dependent weighting scheme to derive optimal estimates of the user coordinates and realistic accuracy measures. The WFN approach, combined with the single epoch (instantaneous) ambiguity resolution algorithm is presented here as one approach to real-time kinematic (RTK) GPS. Since for baselines exceeding ~100 km, the instantaneous ambiguity resolution may not always be possible due to the increasing observation noise and insufficient number of observations to verify the integer selection, an alternative approach, based on a single- (or multiple-) baseline solution, supported by a double-difference (DD) ionospheric delay propagated from the previous epoch is also presented. In this approach, some data accumulation, supported by the network-derived atmospheric corrections, is required at the beginning of the rover data processing to obtain the integer ambiguities; after this initialization period, the processing switches to the instantaneous RTK positioning mode. This paper presents a discussion on the effects of the network geometry, station separation and the data reduction technique on the final quality and reliability of the rover positioning solution. A 24-h data set of August 31, 2003, collected by the Ohio Continuously Operating Reference Station (CORS) network was processed by both techniques under different network geometry and reference station separation. Various solutions, such as (1) single-baseline solution for varying base-rover separation, (2) multi-baseline solution with medium-range base separation (over 100 km), and (3) multi-baseline solution with long-range base separation (up to 377 km), were obtained and compared for accuracy and consistency. The horizontal positioning accuracy achieved in these tests, expressed as the difference between the estimated coordinates and the known rover coordinates, is at the sub-decimeter level for the first approach, and at the centimeter-level for the second method, for baselines over 100 km. In the vertical coordinate, decimeter- and sub-decimeter levels were achieved for the two approaches, respectively. Even though all the results presented here were obtained in post-processing, both algorithms are suitable for real-time applications.