In this paper a novel robust fractional model predictive controller is designed for under-actuated robotic system. The considered under-actuated robotic system is an inverted pendulum on a cart system with two degree of freedom and a control input. The system is modeled to its fractional equivalent model and approximated from the Oustaloup-Recursive-Approximation. From the approximated model a best model is chosen and fractional model predictive controller is designed. Robustness of the fractional model predictive controller is checked to the variations of the system parameters of inverted pendulum on a cart system.