Practical Consideration of Temporal Correlation in RTK GPS Surveying using a Modified Kalman Filter

Many manufacturers of Real-Time Kinematic (RTK) Global Positioning System (GPS) positioning software ignore the temporal correlations found in GPS measurement errors. To understand the statistical value of GPS positions, temporal correlations must be accounted for. A time-differencing Kalman filter is implemented which uses time-differencing to account for temporally correlated errors.;Error ellipses (95% probability) are formed and analyzed in the local coordinate system. When temporal correlation is ignored, far more than 5% of the results produce error ellipses that do not contain the true coordinates. Using the time-differencing modified Kalman filter, the resulting 95% probability error ellipses included the truth position approximately 95 percent of the time without becoming over-pessimistic.;A 3-dimensional survey network was formed on the University of Calgary campus using traditional survey methods. The adjusted network was surveyed repeatedly using commercial RTK equipment using baselines up to 10km. Positions were processed using both commercial RTK software and the modified Kalman filter.