Research On Kalman Filtering Optimized Algorithms In GNSS Navigation

Kalman filter is an efficient and recursive filter,and it could estimate the state parameters of dynamic system from the contaminated observations.In GNSS navigation,Kalman filtering could estimate the carrier's state parameters through dynamic model and observation.But when dynamic model is not accurate or observations contain errors,the result of Kalman filtering would not reflect the real state of carrier.So based on Kalman Filtering,several technologies and improved algorithms are put forward to improve the precision and robustness of GNSS navigation in this paper.Aiming at the dynamic model errors in Kalman Filtering,an improved method of interacting multiple model(IMM)is presented.It contains some single dynamic models,like constant velocity model(CV),constant acceleration model(CA)and current statistical model(CS).It bases on markov hypothesis,and adjusts the transition probabilities between its single models timely.Thus,its ability of tracking target is improved.The traditional Kalman filtering or extended Kalman filtering(EKF)could not estimate nonlinear system faultlessly.The unscented Kalman filtering(UKF)or cubatuer Kalman filtering(CKF)could not run when state covariance is incomplete positive.Also the dynamic model errors make the result dissatisfied.Aiming at solving these problems,an algorithm of strong-tracking reduced square-root CKF is put forward.The experiment results indicate that this algorithm is suitable for strong nonlinear system,and has a better computational efficiency,numerical stability and strong tracking ability.Because prior information can improve the precision of filtering result,three algorithms with constraints based on prior information are presented.One is based on velocity prior information,one is based on road prior information,and the other is based on height prior information.The foundation of velocity constraint is the fact velocity is the first derivative of position about time.It builds the adaptive factor through the coefficient of variation in velocities and makes constraints to the filtering results.And the road constraint builds up a relationship between road prior information and position parameters to get the constraint results directly.And there are two methods that one named pseudo-observation method and the other named projective method used in height constraint.Then analyze the difference between EKF and deterministic sampling filtering when used pseudo-observation method.Finally give an argument for the principle of height-aided positioning by pseudo-observation method when in severe environment and give the way of precision analysis.For solving the problem of getting a reference for judging abnormity when used traditional adaptively robust filterings,two new adaptively robust filtering algorithms are put forward.One is based on observation outlier's detection and reparation,and one is based on receiver autonomous integrity monitoring.The experimental results indicate that the two new adaptively robust filtering algorithms do not rely on a reference to judge abnormity,so they have a wider application and better robustness.Then based on the doppler observation outlier's detection and reparation,a robust least square algorithms for velocity determination is presented.