Cheap decoupled control (Q5947657)

Requiring a controlled system to be decoupled may increase some performance costs, however, decoupling may be desirable from an applied perspective. The main results of the paper are an explicit quantification of decoupling cost in terms of the average quadratic tracking error. Decoupling increases the cheap control cost by spreading non-minimum-phase zeroes and unstable poles. This increase, quantified by the ``decoupling cost factor'', may be small or large, depending on the alignment of zero and pole directions, and the extent of zero and pole spreading. For state feedback, the decoupling cost factor is limited by \(p\), the number of outputs. For output feedback, however, it may be arbitrarily large. The analysis exploits the parametrization of all decoupling controllers, together with Wiener-Hopf frequency domain techniques. Two illustrative examples are considered, one with a small decoupling cost factor, and one with a larger.