Sintonia de controladores baseado na incerteza de identificação de sistemas

Abstract

System identification can provide beyond the model parameters - be it continuous, discrete, linear or nonlinear - a measure of the parameters uncertainty, encoded in the covariance matrix. Thus, when designing a controller for an identified model, we can estimate confidence intervals for step and impulse responses, Bode and Nyquist diagrams, through an uncertainty propagation process. Thus, the performance of the closed-loop controller can be seen as a random variable - with expected value, variance, and confidence intervals. Considering model identification uncertainty, different formulations of controller optimum tuning are proposed, especially the minimization of the ISE’s expectation, for which some analytical results are demonstrated.