Balancing and reacting to strong and unexpected pushes is a critical\nrequirement for humanoid robots. We recently designed a capture point based\napproach which interfaces with a momentum-based torque controller and we\nimplemented and validated it on the iCub humanoid robot. In this work we\nimplement a Receding Horizon control, also known as Model Predictive Control,\nto add the possibility to predict the future evolution of the robot, especially\nthe constraints switching given by the hybrid nature of the system. We prove\nthat the proposed MPC extension makes the step-recovery controller more robust\nand reliable when executing the recovery strategy. Experiments in simulation\nshow the results of the proposed approach.\n