A neural network-based controller for servo drives

This paper outlines the design procedure of a neural network-based controller for tracking control systems. The neural net learns not to identify the inverse dynamics of the plant but to classify the feedback error and its time derivative. Correspondingly, the controller generates the appropriate control signal to cause the servo output to follow a prescribed trajectory. No a priori knowledge of the controlled object dynamics is required, thus enabling the practicality of the proposed method for a variety of servo drive applications. The tracking performance and robustness are verified through the control of an overhead crane model