A partial flatness approach to nonlinear moving horizon estimation

Moving horizon estimators based on an optimization formulation have been proposed as an alternative to extended Kalman filters for constrained nonlinear estimation. An efficient approach to the solution of the nonlinear dynamic optimization resulting from the nonlinear moving horizon estimation (NMHE) problem is presented in this paper. The dynamic optimization problem for the continuous NMHE is transformed into a lower dimensional nonlinear programming problem by eliminating the dynamic constraints for a differentially flat nonlinear system. For the case where the system is not differentially flat, a subset of the nonlinear differential equations can be eliminated. The optimization scheme is demonstrated for the disturbance estimation in a nonlinear chemical reactor.