Self-Tuning Control With Pole-Zero Placement.

Parameter estimation of linear Discrete-time systems by linear programming using two error criteria, namely minimisation of the sum of the absolute values of the errors and minimisation of the maximum absolute error is developed and results are compared with those obtained by the least squares method. The mathematical models are then extended to provide combined structure and parameter identification by mixed integer programming. A method to place both the closed-loop poles and zeros for non-minimum phase systems is developed. The controller can be made self-tuning and its adaptive response remains invariant to changes in process zeros. A method is proposed to vary the system response from exact model-reference following through to a steady state correspondence so that the response with minimum deviation from the reference model can be tuned. A simple variable pole-placement controller is developed and its robustness in the prescence of process-model mismatch and input/output non-linearity is assessed. Two solutions to the implicit selftuning pole-placement regulator problem are proposed and the results extended to develop two variable pole-placement regulators which give stability and the ‘least’ variance of the output in the prescence of process~-model mismatch. The implicit self-tuning generalized minimum variance control of Clarke-Gawthrop is reformulated to give implicit estimation of the control weighting in order to control non-minimum phase systems. It also gives a self adjusting controller which tunes to give least variance of the output when a process-model mismatch exists.