The Extended Innovation Kalman-Sliding Filter for Nonlinear Estimation

Predicting and planning a path and extracting the current location are important aspects in the fields of navigation, localization, and autonomous vehicles. This brief paper belongs to these applications with measurement signals that are obtained from linear sensors. The kinematic states of a vehicle, and the maneuvering angle, are extracted by a filter from a noisy environment. Filters are considered to be either accurate or robust, and typically not both (a trade-off exists). In this paper, we introduce a method that combines accuracy with robustness. The well-known extended Kalman filter (EKF) is combined with the relatively new sliding innovation filter (SIF). The proposed algorithm makes use of the EKF gain and structure while utilizing the robustness of the SIF switching-based gain in an effort to provide a good estimate of the states. The result is a suboptimal nonlinear estimation strategy that resists uncertainties and disturbances. The proposed filter is demonstrated on a vehicle in the Cartesian coordinate while maneuvering and performing turns. The results are compared to the classical EKF and SIF.