Unscented Tobit Kalman filtering for switched nonlinear systems with censored measurement

Abstract

This paper presents a new unscented Tobit Kalman filtering (UTKF) algorithm for switched nonlinear systems with unknown modes and censored measurements. The stochastic switching is considered where the mode information cannot be accessed directly. To reflect the censoring phenomenon, a series of decision variables are introduced and the Type-II Tobit Model is used to characterize the censored measurements where measurements can be transmitted when the decision variables are greater than prescribed thresholds. The aim of this paper is to design a filtering algorithm such that unknown modes, parameters and state are estimated simultaneously over a given finite horizon. By employing the Variational Bayesian (VB) algorithm, several marginal distributions are generated to approximate the joint distribution with the unknown variables. Subsequently, by resorting to the UTKF and Bayesian inference, a new filtering algorithm is developed under the unknown modes and censored measurements. Finally, a simulation result is employed to illustrate the proposed algorithm.