Door Pose Estimation and Robot Positioning for Autonomous Door Opening

Raimund Edlinger, Ulrich Mitterhuber, Roman Franz Froschauer, Andreas Nüchter

Publikation: Beitrag in Buch/Bericht/TagungsbandKonferenzbeitragBegutachtung

50 Downloads (Pure)

Abstract

For autonomous robots to deliver value in human centered environments, they must be able to autonomously open doors. For doing so, they have to overcome multiple challenges, one of which is, to estimate the desired door's orientation and position. The information if the door handle is on the left or right side of the door must also be obtained. In this work, a novel method, solving the stated problems is proposed. It's perception is based on a sensor fusion of a monocular camera based state-of-the-art deep learning object detection algorithm with a 2D laser scan and subsequent line estimation. Additionally, a differential drive controller, using the advancement of continuous goal pose updating, is proposed. During real-world experimentation with a differential drive robot, the implemented system was able to position the robot in front of a door every time with sufficient accuracy and is thus found to solve the stated problem successfully.
OriginalspracheEnglisch
TitelPROCEEDINGS OF THE AUSTRIAN ROBOTICS WORKSHOP 2022
UntertitelRobotics for Assistance and in Healthcare
ErscheinungsortVillach
Seiten36-41
Seitenumfang6
ISBN (elektronisch)978-3-99076-109-0
PublikationsstatusVeröffentlicht - 1 Juli 2022

Schlagwörter

  • Door pose estimation
  • Autonomous mobile robots
  • Pose estimation
  • Deep learning
  • Object Detection

Fingerprint

Untersuchen Sie die Forschungsthemen von „Door Pose Estimation and Robot Positioning for Autonomous Door Opening“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren