Articulated estimation and differential geometry for system identification

摘要

A new point of view in system identification based on a geometrical approach is tackled in this paper. A new robust cost function from a chaining of elementary convex norms (ECN) is proposed. This chain is randomly articulated in order to treat more efficiently natural outliers in data-set during the estimation phase. Estimated parameters can be seen according to a geometrical approach and considered as random fields. Each of them, named articulated estimator random field (AERF) is a manifold or stratum of a stratified space with Riemannian geometry properties. Moreover a system model validation geometric criterion (SYMOVAGEC) based on the expectation of Ricci scalar curvatures of the stratified space for the validation of system model structures Msysis presented. Numerical simulations are provided and discussed to demonstrate the usefulness of the proposed method.