Offline Reinforcement Learning for Quadrotor Control: Overcoming the Ground Effect

Luca Sacchetto, Mathias Korte, Sven Gronauer, Matthias Kissel, Klaus Diepold

Publikation: Beitrag in Buch/Bericht/KonferenzbandKonferenzbeitragBegutachtung

1 Zitat (Scopus)

Abstract

Applying Reinforcement Learning to solve real-world optimization problems presents significant challenges because of the large amount of data normally required. A popular solution is to train the algorithms in a simulation and transfer the weights to the real system. However, sim-to-real approaches are prone to fail when the Reality Gap is too big, e.g. in robotic systems with complex and non-linear dynamics. In this work, we propose the use of Offline Reinforcement Learning as a viable alternative to sim-to-real policy transfer to address such instances. On the example of a small quadrotor, we show that the ground effect causes problems in an otherwise functioning zero-shot sim-to-real framework. Our sim-to-real experiments show that, even with the explicit modelling of the ground effect and the employing of popular transfer techniques, the trained policies fail to capture the physical nuances necessary to perform a real-world take-off maneuver. Contrariwise, we show that state-of-the-art Offline Reinforcement Learning algorithms represent a feasible, reliable and sample efficient alternative in this use case.

OriginalspracheEnglisch
Titel2023 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2023
Herausgeber (Verlag)Institute of Electrical and Electronics Engineers Inc.
Seiten7539-7544
Seitenumfang6
ISBN (elektronisch)9781665491907
DOIs
PublikationsstatusVeröffentlicht - 2023
Veranstaltung2023 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2023 - Detroit, USA/Vereinigte Staaten
Dauer: 1 Okt. 20235 Okt. 2023

Publikationsreihe

NameIEEE International Conference on Intelligent Robots and Systems
ISSN (Print)2153-0858
ISSN (elektronisch)2153-0866

Konferenz

Konferenz2023 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2023
Land/GebietUSA/Vereinigte Staaten
OrtDetroit
Zeitraum1/10/235/10/23

Fingerprint

Untersuchen Sie die Forschungsthemen von „Offline Reinforcement Learning for Quadrotor Control: Overcoming the Ground Effect“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren