Control primitives for fast helicopter maneuvers

Abstract

In this paper, we introduce a framework for learning aggressive maneuvers using dynamic movement primitives (DMP) for helicopters. Our ultimate goal is to combine these DMPs to generate new primitives and demonstrate them on a 3-DOF (3 Degrees of Freedom) helicopter. An observed movement is approximated and regenerated using DMP methods. After learning the movement primitives, the partial contraction theory is used to combine them. We imitate the aggressive maneuvers that are performed by a human and use these primitives to achieve new maneuvers that can fly over an obstacle. Experiments on the Quanser 3-DOF Helicopter demonstrate the effectiveness of our proposed method. In addition, we linearly combine DMPs and propose a new type of DMP. We also analyze Matsuoka's oscillator and Hopf oscillator using contraction theory.