The control of dynamically stable hopping robots has made great progress in the last decades. This paper deals with modeling and control of Kangaroo hopping robot. It is a one-leg planar hopping robot which imitates the mode of displacement of kangaroos. Lagrangian dynamic model of the hopping robot is computed on the different phases of the jumping cycle. A new control scheme is proposed to control the leg thrust during stance phase for planar hopping. A nonlinear model predictive control has been combined with Raibert's approach which has improved significantly its performances. A simulator has been developed to simulate the behavior of the controlled robot. The proposed control approach is validated in simulation and is compared with Raibert's approach.