Article révisé par les pairs
Résumé : This article addresses the design of a nonlinear observer for nonlinear uncertain discrete-time descriptor systems. The nonlinearities that appear in both state and output equations fulfill a Lipschitz-like condition locally. Besides, the parametric uncertainties can affect the state dynamics as well as the input and output channels. The design methodology ensures input-to-state stability of the observer while minimizing an upper bound on a peak-to-peak performance index relating the disturbances/parametric uncertainties and the error in estimating a given performance signal. The observer design is cast as a convex optimization problem involving a set of linear matrix inequalities subject to a linear matrix equality constraint. The effectiveness of the proposed method is illustrated by numerical examples.