Integrating closed-loop identification and control using control energy sensitivity

Abstract

This paper extracts a plant model from an identified closed-loop system model, and proposes an iterative procedure to find high performance controllers. The algorithm uses control energy sensitivity to connect closed-loop modelling and control design. This sensitivity describes the gradient of the consumed control energy with respect to the achieved output covariance. Hence it captures the relative importance of each output channel in the closed-loop system behaviour. Therefore the identification incorporating control energy sensitivity can characterize models to generate a controller for achieving better closed-loop performance. The procedure is demonstrated in a structural control problem.