<p>This work provides a robust model predictive control framework tailored for tracking piece-wise constant reference signals for nonlinear dynamics subject to additive disturbances. The approach integrates setpoint optimization and robust constraint satisfaction into a unified optimization problem, guaranteeing the robust stability within a vicinity of an optimal admissible setpoint. A crucial feature of the approach is its ability to preserve recursive feasibility despite abrupt variations in the target. An offline implementation based on set-valued system representations is also discussed. Numerical examples demonstrate the effectiveness of the controller.</p>