Many practical applications of control system design based on input-ou
tput measurements permit the repeated application of a system identifi
cation procedure operating on closed-loop data together with successiv
e refinements of the designed controller. Here we develop a paradigm f
or such an iterative design. The key to the procedure is to account fo
r evaluated modelling error in the control design and, equally, to let
the closed-loop controller requirements determine the identification
criterion. With an H-2 control problem, this is achieved by frequency
weighting the linear-quadratic Gaussian (LQG) control criterion with f
ilters that reflect the closed-loop plant/model mismatch, and by filte
ring the identifier signals used in a least-squares identification sch
eme in a logical and mutually supportive fashion.