Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Robust Model Predictive Control

Back
Abstract:
A major defect of conventional model predictive control is its `open-loop nature' which causes difficulties when uncertainty in the form of disturbances, state estimation error, or model error is present. With conventional model predictive control, it is impossible to contain the spread of the predicted trajectories so that the optimal control problem (that is solved on-line to yield the current control action) is either impossible to solve or has an unduly conservative solution. In addition, conventional model predictive control loses, in the presence of uncertainty, an important invariance property: the ability to determine a feasible solution to the current optimal control problem from the solution to the prior problem. `Feedback' versions of model predictive control transcend these difficulties but require prohibitively intensive computation. A form of `feedback' model predictive control that overcomes the disadvantages of conventional model predictive control but which has manageable computational complexity is presented. The optimal control problem, solved on-line, yields a `tube' and an associated affine control law that maintains the controlled trajectories in the tube emph{despite} uncertainty. Asymptotic stability of the controlled system is established.

Type of Seminar:
Public Seminar
Speaker:
Prof. David Q. Mayne
Imperial College England
Date/Time:
Jun 04, 2002   17:15
Location:

ETH Zentrum, Gloriastrasse 35, Building ETZ, Room E6
Contact Person:

F. Borrelli
File Download:

Request a copy of this publication.
Biographical Sketch:
Professor Mayne received the BSc and MSc degrees from the University of Witwatersrand, Ph.D and D.Sc (Eng) from the University of London and the honorary degree of Dr.Tech from Lund University. He has had appointments at the University of Witwatersrand, British Thonmson Houston Company, Imperial College, and the University of California, Davis as well as visiting research appointments at Harvard University, University of California, Berkeley and Santa Barbara, and the University of Newcastle, Australia. His research interests include optimization, nonlinear control and model predictive control. He is a Fellow of the Royal Society, a Fellow of the Royal Academy of Engineers, and Fellow of Imperial College.