Self-tuning controller design and implementation

The discretization of the continuous-time plant with respect to the location of the discrete-time zeros of the transfer function and to possible models of typical load-disturbances is described. The general theory of predictive control is shown to have interpretations as Smith predictive, detunable model-reference, and generalized minimum variance laws. Extensions to include feedforward compensation for known disturbances are provided. The pole-placement method is derived to cover the servo and the regulator problem, and in the latter case it is shown that a simple estimation algorithm is sufficient. General recursive estimation methods based on minimizing the prediction error are discussed, particularly numerically stable versions of the recursive least squares method. A simple self-tuner is provided as a combination of an estimator and a basic control law, and the outline of the corresponding program given. Applications of self-tuning are summarized.