Long-range predictive control with a terminal matching condition

Abstract

The long-range predictive control strategy is modified by augmenting the generalized predictive control law with a terminal matching condition. This condition is formulated as a weighting term on the output steady-state error. The final control law only minimizes the square of prediction errors over a small future prediction horizon and at steady state. It is applicable to either transfer function or convolution modeling. For transfer function models, the modified approach reduces the computational load of an adaptive controller by mimicking a large prediction horizon with a smaller one plus a weighting at the steady state. For convolution models, only a few initial step response coefficients and the steady-state gain are required to formulate the new long-range predictive controller. The nonadaptive closed-loop analysis reveals that this new steady-state error weighting term (i) reduces the additional order otherwise required due to inclusion of an integrator for offset elimination, and (ii) provides the stabilizing effect of a large predictive control horizon. A closed-loop adaptive control system based on this modified control law has been successfully applied to the control of mean arterial blood pressure.