Set-Membership Nonlinear Filtering with Second-Order Information

Authors:	Calafiore Giuseppe, Politecnico di Torino, Italy Bona Basilio, Politecnico di Torino, Italy
Topic:	1.1 Modelling, Identification & Signal Processing
Session:	Set-Membership Techniques for Identification
Keywords:	Set-membership filtering, Ellipsoidal bounds, Nonlinear filters, Semidefinite programming

Abstract

In this paper, we develop a numerically efficient scheme for set-membershipprediction and filtering for discrete-time nonlinear systems, that takes intoexplicit account the effects of nonlinearities via local second-order information.The filtering scheme is based on a classical prediction/updaterecursion that requires at each step the solution of a convexsemidefinite optimization problem.The technical results discussed in the paper build uponthe recently developed paradigm of uncertain linear equations (ULE) andsemidefinite relaxations.