CONSTRAINED PREDICTIVE CONTROL AND MODEL WEIGHTING ESTIMATION OF LOW ORDER PLANTS

摘要

The class of control algorithms known as “model predictive control (MPC) techniques” solves a wide variety of control tasks involving constrained optimization of quadratic objective functions. The conventional formulation of MPC based on FIR models assumes that the plant dynamics are known sufficiently well such that the model does not need to be updated online. In this paper, we propose a simple adaptation technique to update the model online when adaptation is necessary, but the plant dynamics are known to vary within a known compact class of models. Input and/or output constraints are handled by a robust optimization algorithm. A numerical example illustrates the application of the proposed scheme.