[G05-1] Indoor and outdoor navigation

[G05-1-06] Improving GNSS RTK and kinematic PPP positioning through extended Kalman filter tuning

Marco Aurelio Moraes de Mendonca, ^○Marcelo C. dos Santos (University of New Brunswick, Fredericton, NB, Canada)

In the past decades, the number of global GNSS data analysis centers and reference stations has consistently increased – from about 100 reference stations in 1995 to more than 500 nowadays. The quality of the orbits, clocks and atmospheric behavior modeling products increased at the same pace. As of February 2017, the final IGS orbits are believed to be accurate at a 2.5 cm level and the clocks at a 75 ps RMS level. Considering as well the increasing quality of GNSS receivers and antennas, often the fine-tuning of positioning estimation for applications demanding high-precision lies in the correct parameterization of the filter. The Extended Kalman Filter (EKF) allows users to set stochastic variations for the parameters to be estimated, such as antenna acceleration, ionospheric, and tropospheric delays random-walk. By under-constraining those parameters with generic values, the error budget is ill-distributed between the different effects on the final estimated position. Also, over-constraining them will force the mathematical model to apportion the same error budget wherever the system allows, often resulting in significant positioning errors. The often overlooked EKF tuning is hereby addressed by a short review of the problem mathematics, followed by experiments. The first dataset is composed of one month of data from the stations UNB3 and UNBJ where different scenarios in RTK and kinematic PPP are analyzed in order to find the specific tuning in a well-known position. The second experiment uses data collected by a GNSS receiver on the top of a vehicle while riding in an urban/suburban environment. The results show 5-10% improvements in accuracy in both RTK and kinematic PPP solutions for the first experiment and indicate a better performance on reconvergence and challenging environments for the second experiment. In conclusion, guiding the EKF with information about the platform and atmosphere can be the difference between achieving or not the desired accuracy.