In this paper the optimal nonlinear predictive cascaded control structure is presented with application to induction motors, which provides global asymptotic tracking of smooth speed and flux trajectories. The controller is based on a finite horizon continuous time minimization of nonlinear predicted tracking errors. With full state measurement assumption, the robustness properties with respect to electrical parameter variations and load disturbance is presented. Finally computer simulations show the flux-speed tracking performances and the disturbance rejection capabilities of the proposed controller in the nominal and mismatched parameters case.