This paper introduces a methodology to design automatically QFT (Quantitative Feedback Theory) robust controllers for SISO (single input-single output) plants with model uncertainty. The method generalizes previous automatic loop-shaping techniques, avoiding restrictive assumptions about the search space. This methodology applies two strategies: a) Evolutionary Algorithms, and b) Genetic Algorithms (GA). In both cases the objective is to search the QFT robust controller that fulfils the control specifications for the whole set of plant models within the uncertainty. Each strategy has been applied to a benchmark in order to validate the techniques.

No abstract

This paper introduces a new methodology to synthesize automatically robust controllers in the Quantitative Feedback Theory (QFT) framework. The method avoids the classical gridding of the controller's phase, and deals with multi-objective specifications and parametric uncertainty in the plant model. By tacking the required robust stability and robust performance specifications, and grouping them into two nonlinear quadratic inequalities, the method derives a nonlinear and frequency-dependent expression for the controller magnitude, which is independent of the controller phase. Then, by evaluating this expression for every frequency of interest, and using a least-square-type algorithm with phase constraints to find the parameters of an a priory fix order controller structure, the method finds automatically the most appropriate controller parameters to meet all the multi-objective specifications for all the plants within the uncertainty. The method is exemplified with a DC motor control application.