In this paper we propose a design methodology for the robust regulation of uncertain nonminimum-phase multivariable linear systems. The method basically consists of defining a suitable vector of parameterized desired output functions and of designing a parameterized decoupling controller of the nominal plant. Then, a parameterized command function for each input is determined by means of a stable dynamic inversion procedure. Eventually, solving an optimization problem permits to minimize the worst-case settling time, subject to amplitude limits on the control variables and constraints on the overshoots and undershoots of the outputs. A worked example based on a three-inputs three-outputs plant shows the effectiveness of the methodology.