The Strong Tracking Innovation Filter

Abstract

Sliding innovation filter (SIF) has recently been introduced as a robust strategy for estimation of linear systems. The SIF has been extended to nonlinear systems via analytical linearization. However, as the performance of the extended SIF (ESIF) degrades in the presence of severe nonlinearities, this article has initially developed a derivative-free cubature SIF (CSIF) that uses statistical linearization for the error propagation. In addition, the SIF gain has been reformed to incorporate the innovation covariance matrix, thus reducing the estimation error. Furthermore, the adaptive fading factor has been employed to strengthen the robustness and convergence properties of the CSIF against abrupt changes of state variables. Simulation results of the proposed estimation algorithm named the strong tracking innovation filter (STIF) have been compared to those of the ESIF and the CSIF in different conditions of the modeling error in the dynamic system and statistical characteristics of the system inputs. This comparison has demonstrated the superior performance of the STIF in terms of the convergence rate, estimation accuracy, and the chattering elimination. Integration of the STIF into a sliding mode controller for the concurrent estimation and control of a Mars lander has reconfirmed the robustness and accuracy of the proposed STIF.