---
title: "Deep Drone Acrobatics (Extended Abstract)"
authors:
  - "Elia Kaufmann"
  - "Antonio Loquercio"
  - "Rene Ranftl"
  - "Matthias Müller"
  - "Vladlen Koltun"
  - "Davide Scaramuzza"
doi: "10.24963/ijcai.2021/650"
url: "https://ijcai.org/proceedings/2021/650"
pdf: "https://ijcai.org/proceedings/2021/0650.pdf"
venue: "Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence"
year: 2021
pages: "4780-4783"
track: "Sister Conferences Best Papers"
keywords:
  - "Robotics: Learning in Robotics"
  - "Machine Learning Applications: Applications of Supervised Learning"
  - "Robotics: Behavior and Control"
---

# Deep Drone Acrobatics (Extended Abstract)

Elia Kaufmann, Antonio Loquercio, Rene Ranftl, Matthias Müller, Vladlen Koltun, Davide Scaramuzza

*Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence*, pages 4780-4783. DOI: [10.24963/ijcai.2021/650](https://doi.org/10.24963/ijcai.2021/650)

## Abstract

Acrobatic flight with quadrotors is extremely challenging. Maneuvers such as the loop, matty flip, or barrel roll require high thrust and extreme angular accelerations that push the platform to its limits.  Human drone pilots require years of practice to safely master such maneuvers. Yet, a tiny mistake could make the platform lose control, and brutally crash. This short paper describes an approach to safely train acrobatic controllers in simulation and deploy them with no fine-tuning zero-shot transfer on physical quadrotors. The approach uses only onboard sensing and computation.