A robust extended Kalman filtering for linearization errors

In this paper, we propose a new robust design of the extended Kalman filter(EKF) for nonlinear state estimation problems. In order to give the robustness to the conventional EKF, we consider the linearization error instead of neglecting the nonlinear higher order terms. The linearization errors are represented as a linear function of estimation error and is treated as model uncertainty in linear model. Additionally, we propose the systematic technique for predicting the linearization errors by using the current estimated state and one step ahead one. And, the linearized model and prediction of the linearization errors can be obtained within the Kalman filtering framework.