Fixed order controller synthesis for electro-mechanical servo-systems

Project members: C.W.J. Hol, C.W. Scherer

Sponsored by: Philips CFT

The demanding requirements on speed and accuracy on electro-mechanical servo-systems imply the use of multi-variable feedback control based on a dynamic model of the system. A state-of-the-art method for multi-variable controller synthesis is through optimization.

One of the main disadvantages of controller synthesis is the high order (McMillan degree) of the resulting controller which equals the order of the dynamical model plus the order of the weighting functions, typically in the range of about one hundred. In real-time implementations on electro-mechanical systems with a very high sampling rate (typically around 1kHz), the computation of the controller action becomes more expensive with increasing controller order. This reveals the need for a low-order controller synthesis method which goes beyond existing techniques in being applicable to systems of order larger than fifty.

This research project aims at developing a synthesis method to compute fixed-order controllers for higher-order models of electro-mechanical servo-systems. The algorithm must be implemented in a numerically stable fashion to facilitate the dissemination of reduced order synthesis in the industrial practice.