Introducing Caution in Iterative Controller Design

Abstract

Iterative identification and control design extend classical design methods by using on-line experimental performance data to adjust the feedback regulator. The inclusion of this new information concerning achieved controller properties permits a focus on actual performance, as opposed to designed performance. In the context of Robust Control, this is a fundamentally important piece of information. However, controller adjustment for performance improvement needs to be balanced against closed-loop stability requirements. To achieve this balance it is critical not to modify the controller too outlandishly in any one step but, rather, to apply some cautious updating procedure to limit the change to the controller. Likewise, alterations to the desired controlled performance need to be made gradually rather than dramatically. Our approach in this paper is to explore the origin of this caution and to present some methods derived and applied in a practical problem of sugar-cane crushing mill control.