Algorithm to compute nonlinear partial observer normal form with multiple outputs

Wided Saadi (1) , Driss Boutat (2) , Gang Zheng (3) , Lassaad Sbita (1) , Lei Yu (4)

1 ENIG - Ecole Nationale d'Ingénieurs de Gabès
2 PRISME - Laboratoire pluridisciplinaire de recherche en ingénierie des systèmes, mécanique et énergétique
3 DEFROST - Deformable Robots Simulation Team
4 Wuhan University [China]

It is well-known that observer design is a powerful tool to estimate the states of a dynamical system. Given a multi-output nonlinear dynamical system whose states are partially observable, this paper investigates the problem of observer design to estimate those observable states. It considers firstly a nonlinear system without inputs, and provides a set of geometric conditions, guaranteeing the existence of a change of coordinates which splits the studied nonlinear dynamical system into two subsystems, where one of them is of the well-known nonlinear observer normal form, for which a Luenberger-like observer can be designed. An extension to nonlinear systems with inputs has then been deduced.

Mots clés: Partially observable; observer design; Differential geometric methods

Domaines: Automatique

Wided Saadi, Driss Boutat, Gang Zheng, Lassaad Sbita, Lei Yu. Algorithm to compute nonlinear partial observer normal form with multiple outputs. IEEE Transactions on Automatic Control, 2020, 65 (6), pp.2700 - 2707. ⟨10.1109/TAC.2019.2946528⟩. ⟨hal-02406825⟩