A tracking problem is considered for a Wiener model. A two-layer hierarchical control structure is designed: the upper level controller operates at a slow rate and computes the inputs to be ideally provided to the system; the effective control actions are provided by actuators placed at the lower layer and having faster dynamics. Model predictive control (MPC) laws are synthesized for both layers. In order to cope with the discrepancy between the ideal and the effective control action, a robust MPC controller is designed at the upper level. Such a controller can switch among different operating conditions ensuring different level of robustness. In doing so, the overall controller guarantees steady-state zero error regulation for constant reference signals while trading off robustness versus system performance. A numerical example is reported. Copyright © 2011 John Wiley & Sons, Ltd.