AbstractTwo auto-tuning controllers, one based on correlation identification technique and the other on process reaction curve method, are proposed and compared. The underlying controller is a predictive PI controller, wliich is an enhanced PI controller. For correlation technique, an online algorithm is developed for identifying the impulse response and hence determining the open-loop step response. For process reaction curve method, an online procedure is used to obtain a first-order-with-tiine-delay approximation to multistaged higher-order systems. Based on the information obtained, the parameters of the controller are updated. Simulation studies compare the per-formance of each algorithm in the presence of noise and time-varying systems.