New approach to analysis and design of Smith-predictor controllers

Abstract

An equivalent representation of the Smith-predictor controller is presented and it provides a new framework for analysis and design of the Smith systems. In particular, the single-loop controller can be viewed as a special mismatched Smith system. On this platform, process predictor selection can be posed as an optimization problem such that the compensated process consisting of the process and optimal predictor yields the best achievable performance. As a result, it is observed that, in contrast to intuition, a mismatched Smith system may give a better performance over the perfectly matched one. Examples are included for illustration.