The advantages and disadvantages of kalman filtering in iterative learning control
Conference proceedings article
Authors/Editors
Strategic Research Themes
No matching items found.
Publication Details
Author list: Panomruttanarug B., Longman R.W.
Publication year: 2008
Volume number: 130 PART 1
Start page: 347
End page: 365
Number of pages: 19
ISBN: 9780877035442
ISSN: 0065-3438
eISSN: 0065-3438
Languages: English-Great Britain (EN-GB)
Abstract
Iterative learning control (ILC) can eliminate deterministic tracking errors of a control system that repeatedly performs the same tracking maneuver. It iterates with the real world rather than with a model aiming to achieve this zero error, and hence it can achieve lower final error levels in the real world than one could obtain when limiting oneself to using a mathematical model aiming to achieve zero tracking error. There are spacecraft applications when fine pointing sensors perform repeated scans. When there is substantial plant and measurement noise, it is natural to consider using a Kalman filter to improve the signal used by the learning control law. In a separate work, this was analyzed for the sister field of repetitive control (RC). It was observed that when the model in the Kalman filter is imperfect, the use of a Kalman filter can make the final error levels worse rather than better, and care must be taken to assess the benefit in reduced random variations compared to the deterministic error introduced. This is also the case when there are unmodeled repeating external disturbances. Here we show that the same situation applies to ILC if one uses a Kalman filter running in time steps during each iteration. But one can reformulate the Kalman filter to operate in the repetition or iteration domain, and when this is done, one can still converge to zero expected value of the tracking error in the presence of model errors and external repeating disturbances.
Keywords
No matching items found.
Contact Information