Learning from Demonstration (LfD) allows robots to learn tasks directly from humans, but surface interactions on 3D objects present challenges. This paper introduces ProSIP, a method that enables robots to learn surface paths and local features. A specialized tool captures and executes human demonstrations easily. ProSIP works independently of time and on different robots with Cartesian control. The method is demonstrated on an edge-cleaning task for sinks, showing its adaptability to various object shapes and distorted surfaces. Simulations and experiments confirm the approach’s effectiveness.
00:00 Intro
00:10 Recording demonstration data only on front edge
00:23 Cleaning different edges on different sinks
01:05 Outro
Publication:
https://www.acin.tuwien.ac.at/file/publications/cds/pre_post_print/unger2024.pdf
Authors:
Christoph Unger
Christian Hartl-Nesic, https://orcid.org/0000-0003-3054-9435
Minh Nhat Vu, https://orcid.org/0000-0003-0692-8830
Andreas Kugi, https://orcid.org/0000-0001-7995-1690
Industrial Robotics, ACIN, TU Wien
https://www.acin.tuwien.ac.at/industrial-robotics/
LinkedIn: Automation and Control Institute ACIN TU Wien