In this project, it is desired to learn different acrobatic maneuvers through a set of demonstrations shown by an expert pilot. Due to the complexity of the vehicle's dynamics, at first step, it is necessary to find an appropriate set of states that can best represent an acrobatic maneuver. Next, a change of frame of reference from the East-North-Up Coordinates System to Aircraft-Body Coordinates System is applied on the whole demonstration dataset to give a more accurate definition of the maneuver and to handle the problem associated with different starting positions and orientations. To increase the performance of the algorithm, the demonstration data points are filtered and refined. After data preprocessing, the whole motion is encoded using Gaussian Mixture, and finally, an analysis of the model performance is made together with a discussion on the ways in which such a model could be used to control the aircraft.