Skip to main navigation Skip to search Skip to main content

Quadrotor Aerobatic Maneuver Attitude Controller based on Reinforcement Learning

  • Beihang University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

A model-free attitude controller design framework for aerobatic maneuver of quadrotor is proposed in this paper. We utilize Proximal Policy Optimization, a reinforcement learning algorithm to train a neural network controller. The proposed controller can handle the highly coupled nonlinearity of aerobatic maneuver dynamic while requires no explicit dynamic model of the quadrotor. Compared with traditional PID controller, the proposed controller shows advantage in both rapidity and overshoot when tracking aerobatic maneuver attitude commands.

Original languageEnglish
Title of host publicationASCC 2022 - 2022 13th Asian Control Conference, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages2450-2453
Number of pages4
ISBN (Electronic)9788993215236
DOIs
StatePublished - 2022
Event13th Asian Control Conference, ASCC 2022 - Jeju, Korea, Republic of
Duration: 4 May 20227 May 2022

Publication series

NameASCC 2022 - 2022 13th Asian Control Conference, Proceedings

Conference

Conference13th Asian Control Conference, ASCC 2022
Country/TerritoryKorea, Republic of
CityJeju
Period4/05/227/05/22

Keywords

  • Neural Network based control
  • Nonlinear control
  • Reinforcement learning

Fingerprint

Dive into the research topics of 'Quadrotor Aerobatic Maneuver Attitude Controller based on Reinforcement Learning'. Together they form a unique fingerprint.

Cite this