Motion Primitives for Robotic Flight Control

by B.E. Perk & J. J. E. Slotine arXiv.org E-print Archive, 25 Sep 2006 We introduce a simple framework for learning aggressive maneuvers in robotic flight control. Standard movement primitive techniques are analyzed and extended using nonlinear contraction theory. Accordingly, dynamic primitives, approximated and regenerated from an observed movement, can be stably combined or concatenated for various purposes. We demonstrate our results experimentally on the Quanser Helicopter, in which we first imitate aggressive maneuvers performed by a human, and then use these primitives to achieve new maneuvers that can pass an obstacle. Read more