Robust tuning of PID regulators based on step-response identification

Robust control design requires an accurate knowledge of the uncertainty bound around the nominal model. While classical identification theory qualifies the uncertainty in identified models through confidence regions of the estimated parameters, recent works have studied frequency-domain error bounds for models identified via least-squares like techniques. In this paper a quantification of errors in models identified through step-response methods is proposed. A model error bound is derived based upon an elaboration of the step response of the process. This bound is then used to formulate simple rules for a robust tuning of PID controllers. A simulation example is also presented, where a comparison of the closed-loop performance obtained with the proposed tuning rules and with other common tuning rules is shown.