Iterative learning control, with applications to a wafer-stage

Iterative learning control, with applications to a wafer-stage

Dijkstra, B.G.

Bosgra, O.H. (promotor)

Design, Engineering and Production

2004-09-08

In the past Iterative Learning Control has been shown to be a method that can easily achieve extremely high performance with very limited effort from the user. Unfortunately an Iterative Learning Controller (ILC) exhibits some side effects such as amplification of system noise and control activity when the system should be at rest. This had led to a primary and secondary objective for this research: -Develop a systematic design method for robustly stable Iterative Learning Control that deals with non-repetitive signals (like noise). -Explore the relations between Iterative Learning Control and trajectory design. To this end a generalized setting has been presented that encompasses all forms of Iterative Learning Control, describing it as a feedback system operation from one trial to the next. This trial-domain feedback setting has served as basis for the development of several finite time ILC design techniques based on classical feedback theory, such as Linear Quadratic optimal control, Kalman Filter designs and multi-objective control. Downside to these methods is the very large order of the problems for ILC, which has been solved in the development of a low-order ILC setting based on a Hamiltonian system description for ILC. This low order setting has been used for the design of time-weighted ILC, allowing for the design of ILC for point-to-point control, resulting in very short settling times and eliminating the presence of control actuation after the trajectory, which is commonly encountered with regular ILC. The low-order finite time ILC design setting has been proven to be very efficient and flexible, which has been illustrated with application to different wafer-stages.

ilc
wafer-stage
input-shaping

http://resolver.tudelft.nl/uuid:b29a7fa2-6bee-4b59-8aaf-b7b016e1eeba

90-370-0209-9

Institutional Repository

doctoral thesis

(c) 2004 B.G. Dijkstra