Model-based control approaches require the mathematical model of system to be controlled, which have been widely used in many applications. However, with the increasing complexity of system dynamics, modeling process by first principles has become more difficult. When the model is inaccurate, model-based control methods would lose the utility.  As data is becoming more readily available, learning from data has then tracked more and more attentions in the control community.  In our research group, we develop a new data-driven method for nonlinear system, which will be implemented on the different robotic platforms.