AbstractA new constrained self‐tuning PID controller is developed based on Clarke and Gawthrop's (1975) generalized minimum variance (GMV) controller. Convergence properties for the new algorithm are derived and the effect of the constraint parameter is evaluated. Extension of the controller to include more than the usual three PID parameters occurs naturally. No process model information is required for our algorithms except the dead time. This is an improvement over other algorithms which impose severe constraints on the process structure.The controller algorithm is tested on a variety of models using simulation as well as under experimental conditions. The algorithm is easy to implement and compares favourably with the more complex GMV controlle