Parametric Identification of Nonlinear Systems: Guaranteed Confidence Regions

In this paper we consider the problem of constructing confidence regions for the parameters of nonlinear dynamical systems. The proposed method makes use of higher order statistics and extends a previous algorithm in [3]. The obtained confidence regions are valid for any finite number of data samples and they are nonconservative, in the sense that they contain the true parameter value with an exact probability. The usefulness of the proposed approach is illustrated in simulation examples. The results presented here are preliminary results from an ongoing research on finite sample properties in nonlinear system identification.