Linear Iterative Learning Control (ILC) algorithms have been known to also perform well for nonlinear systems whose dominant system dynamics are linear. In order for the learning filter to take advantage of more system information, we propose here a model based ILC algorithm which uses an iteration varying learning filter. Before the next iteration’s feedforward control is computed, the linearized system model is first estimated using a least squares approximation. We implement this algorithm on a wafer stage prototype whose dominant system dynamics are linear with a weak nonlinear actuator disturbance. Because the nonlinear disturbance is state dependent, the linear dynamics will shift as the ILC algorithm is converging. We show that as the system converges to the desired trajectory the plant parameter also converge.

This content is only available via PDF.
You do not currently have access to this content.