A self-optimizing non-linear control system

A general technique of self-adaptation is presented which adjusts one parameter of a control system to obtain the best response as determined by some built-in optimizing criterion. The method has been applied to a position-control servo mechanism in which the damping factor is automatically adjusted to give minimum mean squared error. The response of the control system is considered for repetitive step functions and sinusoidal and random signal inputs. The addition of the self-adjusting feature requires little knowledge of the controlled process dynamics or applied signal characteristics. The optimization mechanism compensates for variations in system parameters, non-linear effects and changes in the applied signal. It is shown that the introduction of a self-optimizing element effects a substantial improvement in system performance. Experimental results using an analogue computer are in good agreement with theoretical predictions.