The goal of this project is to push the state-of-the-art in autonomous helicopter flight: extreme aerobatics under computer control.
Stanford’s autonomous helicopter group has made some impressive advancements in the field of autonomous helicopter control, including inverted hovering and performing aerobatic stunts.
The group uses reinforcement learning to teach its control system various maneuvers and has been very successful in doing so. One of their latest achievements wasteaching the bot the emergency landing technique autorotation. Autorotation is used when a helicopter’s engine fails or is disengaged and works by changing the collective pitch to use the airflow from descent to rotate the blades