Robust Extended Kalman Filter with considering all random effects for integrated navigation

In this contribution, a robust prediction is developed to solve a general nonlinear dynamic model with all random effects (GNDRE), which has been defined recently as a fairly general dynamic model. Compared to the general extended Kalman filter (GEKF), the robust GEKF (RGEKF) can solve the GNDRE model when the outliers may exist in observation equations and/or system equations. The developed filter is formulated by condition equations, which leads to obtaining the residuals directly. All of the dynamic problems can be considered as a special case of this model. The new algorithm is implemented in the integrated navigation of low-cost tightly coupled GNSS/Vision/INS systems in the presence of outliers to demonstrate the efficiency of the proposed approach numerically. The average improvement for position is about , for position and 0.99∘ for attitude, in contrast to the EKF algorithm, while these improvements are 0.641m for position and 0.576∘ for attitude compared to the GEKF method.