A Novel Data-Driven Filtering Algorithm for a Class of Discrete-Time Nonlinear Systems

Abstract

Data-driven filtering technique has immense potential and gained significant attention lately. This paper investigates a novel data-driven filtering algorithm based on a new dynamic linearization technique in the framework of Kalman Filter for a class of discrete-time nonlinear systems. Compared with the conventional nonlinear filtering algorithms, such as Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF), the proposed data-driven filtering (DDF) method can not only be applied for nonlinear systems without precise mathematical model or linearization approximation, but also be designed by merely utilizing the I/O measurement data of the plant. The theoretical analysis shows that the proposed approach guarantees uniform ultimate boundedness of the filtering errors. The comparison numerical simulation results verify the effectiveness of the proposed approach.