Real time dynamic precise point positioning with mixture filtering

The linearisation of non-linear observation equations in precise point positioning (PPP) can always bring out additional errors. To weaken such linearisation errors and improve the performance of unscented Kalman filtering (UKF), a kind of mixture filtering based on combined UKF and extended Kalman filtering (EKF) is presented in this contribution. First, the PPP solutions estimated on the base of EKF are taken as predicted information for UKF, which can avoid the impossibility of matrix decomposition during UKF implementation; second, the accuracy of PPP solutions can be optimised by rerunning the UKF. An airborne test is used to demonstrate the performance of the new mixture filtering concept. The experimental results show that the accuracy of dynamic PPP with mixture filtering is better than their counterparts as estimated with EKF only.