Observer design for sampled-data nonlinear systems via approximate discrete-time models

Since exact discrete-time models are often unavailable for nonlinear sampled-data systems, the most realistic approach for observer design is to employ approximate discrete-time models. This paper investigates under what conditions, and in what sense, such an approximate design achieves convergence for the unknown exact discrete-time model. We first present examples which show that designs via approximate discrete-time models may indeed lead to instability when implemented on the exact model. We then present conditions for approximate designs that guarantee robustness for the exact discrete-time model.