Real-time experiments in neural identification and control of disturbance

Designs have been proposed for neural identification and control of dynamic systems with unknown disturbance. These neural designs have been implemented in real-time experiments using a DC motor system with friction as the plant and additive disturbance at the input. The experimental results reveal the capability of neural networks to identify bounded, unknown, periodic disturbance. It also shows the improved performance of the dynamic neural control design as compared to other schemes found in the literature. This control design incorporates a better gradient estimation scheme using a disturbance identification network