This semiplenary talk at the IFAC Symposium on Nonlinear Control Systems (NOLCOS, Stuttgart, September 2004) is proposing state reconstructors for nonlinear systems. Our techniques extend a previous work on state reconstructors for linear systems by the same authors (Reconstructeurs d'état, C.R. Acad. Sci. Paris, série I, 338, 2004, pp. 91-96), which bypasses some of the classic difficulties related to asymptotic observers and Kalman filtering (lack of robustness and knowledge of the statistics). Our viewpoint which avoids the integration of differential equations and therefore any asymptotic estimation yields fast implementable algebraic formulae. Two concrete case studies are presented, which are (differentially) flat. Our state estimation permits a state-feedback around the flatness-based reference trajectory. Convincing computer simulations are provided, which demonstrate the robustness of our control strategy with respect to noises with unknown statistical properties.