Gaussian Mixture Model for 3-DoF orientations

We validate the proposed method in three different applications, learning a 3-dimensional rotation matrix, learning reachable space of a robot, and learning the motion model from demonstrations. We show good experimental results compared to the state-of-the-art method.