Robust state estimation using desensitized Divided Difference Filter

Abstract

This paper develops a robust divided difference filtering approach based on the concept of Desensitized Kalman Filtering. The filters are formulated using a minimum variance cost function, augmented with a penalty function consisting of a weighted norm of the state sensitivities. Solutions are provided for first and second-order Divided Difference Filters. The resulting filters are non-minimum variance but exhibit reduced sensitivity to deviations in the assumed plant model parameters. The proposed algorithms are demonstrated using Monte Carlo simulation techniques for an induction motor state estimation problem with parameter uncertainties.