Robust stability of model predictive control

The closed loop stability of model predictive controllers (MPC) is analyzed in the presence of model plant mismatch (MPM). A "dual-model" state space formulation is used to describe the process and matrix perturbation plus root locus techniques are used to determine robust stability bounds for MPC. The special structure of the model simplifies design and analysis in the presence of mismatch in the gain, time delay, slow dynamics or fast dynamics. Simple examples show that contrary to the expectations of most users, detuning SISO MPC using widely accepted tuning guidelines does not necessarily result in greater robustness, e.g., increasing the control horizon, M, can increase robustness, increasing the prediction horizon P can cause instability.