Date: Wednesday, April 10, 2013
Location: 4088 East Hall (3:10 PM to 4:00 PM)
Title: Linear Model Order Reduction In Model Predictive Control
Abstract: Model order reduction (MOR) is a common task for model-based control design, specifically for those where a copy of the plant model appears explicitly in the resulting controller. This is done to reduce the number of states and thereby the complexity because plant model complexity dictates the controller complexity and robustness. For computationally intensive optimization-based control, such as model predictive control (MPC), MOR is essential to ensure that the controller is implementable on hardware, especially for applications where the computational resources, such as memory and power, are limited and fast sampling is required.
The problem is that most MOR techniques are meant to be applied to a class of model - stable, unstable, passive, dissipative, etc. - without consideration to the control formulation.
This talk will consist of a brief overview of MPC, the techniques and theoretical tools of linear MOR, and then a novel reduction method that accounts for the MPC control formulation.
Speaker: Richard B. Choroszucha
Institution: University of Michigan
Event Organizer: Bowei Wu boweiwu@umich.edu
|
