A systematic method to enhance the robustness of stabilising receding-horizon predictive controllers

Abstract

Although stability guarantees are available for predictive controllers, closed-loop instability may arise due to system uncertainty. The T-design and the Q-parametrisation methods are often used to enhance the robustness of predictive controllers. The former is heuristic, whereas optimisation rules exist for the latter, which makes it systematic. However, the T-design tends to producing larger robustness margins. The T-optimisation presented here is a systematic procedure to choose T based on optimising a quadratic criterion on robustness and thus, overcomes the drawbacks of heuristic approaches. This method can lead to robustness margins larger than those obtained with other techniques.