M.S. Thesis Presentation by Michael Mashner
Wednesday, November 13, 2002

(Dr. Nader Sadegh, advisor)

"Delayed Measurement Based Multirate Kalman Filtering and State Feedback Control"

Abstract

A vision system can be an invaluable tool for machine end affecter position measurement. It can give accurate data of absolute position that is free of system noise and vibration that can greatly improve estimates of the system states. The drawback to the use of vision systems for feedback control is the long image processing time and thus slow sampling rate. A standard method to account for the time delay and sparseness of measurements is to formulate a time varying system with many extra states. This greatly increases the number of online computations to be performed by the controller. A method has been developed to use a Kalman Filter that accounts for the delayed, periodic nature of the vision measurements without significantly increasing computation intensity using a downsampling technique. Noise discretization and noise down sampling will also be investigated as they are an important aspect of Kalman filtering. A multirate control scheme has also been developed to allow easy integration of this Kalman filter into industrial applications.