## 3-Iterative Learning Control

### Mark Bedillion (Carnegie Mellon University)

Friday, October 7, 2022
7:00AM – 11:00AM (PDT USA)

Many precision positioning applications feature trajectories that are repetitively followed, e.g. raster scanning in an AFM. This tutorial introduces the tools and techniques used to design iterative learning controllers (ILC) to substantially reduce errors in tracking such repeated reference signals.  The tutorial will begin by presenting the two fundamental analysis frameworks for ILC (time domain and frequency domain) and present classical results on ILC stability (does the learning converge?) and steady state error (how small does the error get?).  We will also spend time discussing block diagrams and what is meant by the “plant” for ILC systems.

Next ILC design techniques will be described, including PID-type algorithms, plant inversion approaches, $H_\infty$ / $\mu$-synthesis formulations, and linear quadratic ILC.  Each design method will be briefly introduced along with a Matlab-based performance demonstration on a benchmark problem.  With the primary approaches established, the tutorial will move on diving more deeply into the linear quadratic ILC approach to discuss tuning, stability / error properties, and extensions.

The tutorial will end by describing practical implementation issues for ILC controllers, including the impacts of saturation, data storage, and system nonlinearities.

This is a somewhat advanced class in feedback control systems, and it is expected that the attendees will have a basic understanding of linear algebra and frequency domain tuning techniques.

Mark Bedillion is a Teaching Professor in the Mechanical Engineering Department at Carnegie Mellon University who specializes in feedback control systems and mechatronics. Prior to joining the faculty at Carnegie Mellon his academic experience was as an associate professor at the South Dakota School of Mines and Technology. His primary industrial experience comes from a seven-year career at Seagate Technology, where he worked on research and development of servo-control architectures.

Professor Bedillion’s teaching interests include mechatronics and control systems courses at both the undergraduate and graduate levels. His research interests include control for precision positioning systems, STEM education, distributed manipulation, control applications in data storage, and control applications in manufacturing.