Automatic design of multi-variable quantitative feedback theory control systems via evolutionary computation

Abstract

This paper proposes a multi-objective evolutionary automated design methodology for multi-variable quantitative feedback theory (QFT) control systems. Unlike existing analytical and convex optimization-based QFT design approaches, the evolutionary 'intelligent' technique is capable of automatically evolving both the nominal controller and the pre-filter simultaneously to meet the usually conflicting multiple performance requirements in QFT, without going through the sequential and conservative design stages for each of the multi-variable subsystems. In addition, it avoids the need of manual QFT bound computation and trial-and-error loop-shaping design procedures, which are particularly useful for multi-variable or unstable plants where stabilizing controllers may be difficult to synthesize. The effectiveness of the proposed QFT design methodology is validated upon a benchmark multi-variable system, which offers a set of low-order Pareto optimal controllers satisfying all closed-loop performances under practical constraints.