The problem of designing robust inverse-based controllers for multivariable ill-conditioned processes is addressed in this paper. A methodology is developed aimed at finding modified models that, used in inverse-based controllers, are able to make the overall control structure robust. The models are obtained through the solution of an optimization problem which contains in the objective function terms related both to nominal performance and to robustness to uncertainties. A basic advantage of the procedure is that its results can be used in any kind of inverse-based controller. The application to the design of a decoupler and of an internal model controller for a structurally ill-conditioned distillation column is presented to show the effectiveness of the method.