ZSM-IMM and ZPRM-IMM: Two novel interacting multiple model based state estimation algorithms for stochastic switching systems under unknown but bounded noise.

Abstract

In this study, two novel state-estimation algorithms based on an interacting multiple model (IMM) are proposed for stochastically switched linear systems. First, a zonotopic filter is derived from segment minimization. Then, the zonotopic segment minimization based IMM algorithm is proposed, which comprises four steps: input interaction, segment minimization filtering, model probability updating, and output fusion. In addition, to avoid the upper and lower bounds of the zonotope at individual moments from wrapping around the true value, the zonotopic P-radius minimization based IMM algorithm is also studied, which creatively employs the P-radius minimization process in the zonotopic updating step and converts it into a linear matrix inequalities problem. Finally, the two proposed algorithms are verified by numerical simulation and an experimental analysis of a buck-boost circuit.