A simple continuous process is controlled by a digital computer. The process is initially identified and thereafter represented on a digital computer by means of a dynamic model. The optimum response of the model is obtained by adjusting the parameters of the digital controller to minimise a chosen cost function. Once the optimum controller has been determined by methods involving conjugate gradients, as compared to the method of steepest descent, the model is replaced by the actual process, which can then be controlled in an optimum fashion. It is possible for the computer to determine the model of the process and the optimum controller in an uninterrupted manner. The system is subjected to a step demand and a smoothed pseudorandom binary sequence.