Output Feedback Reference Governor for Nonlinear Systems

This paper develops a new reference management algorithm for constrained nonlinear systems with unmeasurable states. Instead of the state itself, the present method utilizes an ellipsoidal region in which the state is guaranteed to lie. Such a region can be obtained by using the set-valued observer due to Scholte and Campbell [9]. The present method requires a solution of the optimization problem which may not be solved effectively. For ease of implementation, we introduce a relaxed problem which is always efficiently solvable. When neither noise nor disturbance are present and the reference is constant, we show sufficient conditions for the modified reference to be settled to the reference in a finite time, and consequently for the convergence of the state to the desired equilibrium. In the presence of noise and/or disturbance, we derive somewhat conservative conditions for the finite-time settling of the modified reference to the original one and for the convergence of the state to the neighborhood of the equilibrium. The effectiveness of the present method is demonstrated by a numerical example.