University of Minnesota
Aerospace Engineering and Mechanics
Spring 2003 Seminar Series



Iterative Learning and Repetitive Control


Prof. Richard Longman

Department of Mechanical Engineering, Columbia University


Abstract

Iterative learning control is a rather new field that develops controllers that learn from previous experience performing a specific command, and try to converge on zero tracking error. An overview of the concepts used for learning as well as experimental results will be given, including applications to robotics, computer disk drives, vibration cancellation in spacecraft, copy machines, particle accelerators, etc. On a commercial robot, the RMS of the tracking error of all joints doing a high speed maneuver was decreased by a factor of 1000 in only 12 repetitions for learning. This is actually below the reproducibility level of the hardware when measured from day to day -- so the methods learn to eliminate tracking errors that are the size of the difference in performance of the robot from one day to the next.

Friday, May 9, 2003
209 Akerman Hall
2:30-3:30 p.m.


Refreshments served after the seminar in 227 Akerman Hall.
Disability accomodations provided upon request.
Contact the AEM Office, 625-8000.