Improving the Generalizability of Robot Assembly Tasks Learned from Demonstration via CNN-based Segmentation

Inigo Iturrate*, Etienne Roberge, Esben Hallundbak Ostergaard, Vincent Duchaine, Thiusius Rajeeth Savarimuthu

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

Kinesthetic teaching and Dynamic Movement Primitives (DMPs) enable fast and adaptable learning of robot tasks based on a human demonstration. A task encoded as a dynamic movement primitive can be reused with a different goal position, albeit with a resulting distortion in the approach trajectory with regards to the original task. While this is sufficient for some robotic applications, the accuracy requirements for assembly tasks in an industrial context, where tolerances are tight and workpieces are small, is much higher. In such a context it is also preferable to keep the number of demonstrations and of external sensors low. Our approach relies on a single demonstration and a single force-torque sensor at the robot tool. We make use of a Convolutional Neural Network (CNN) trained on the force-torque sensor data to segment the task into several movement primitives for the different phases: pickup - approach - insertion - retraction, allowing us to achieve better positional accuracy when generalizing the task primitives to new targets. To the best of our awareness, we are the first to utilize a CNN as a segmentation tool to improve the generalization performance of DMPs.

Original languageEnglish
Title of host publication2019 IEEE 15th International Conference on Automation Science and Engineering, CASE 2019
PublisherIEEE
Publication date2019
Pages553-560
ISBN (Print)978-1-7281-0357-0
ISBN (Electronic)9781728103556, 978-1-7281-0356-3
DOIs
Publication statusPublished - 2019
Event15th IEEE International Conference on Automation Science and Engineering, CASE 2019 - Vancouver, Canada
Duration: 22. Aug 201926. Aug 2019

Conference

Conference15th IEEE International Conference on Automation Science and Engineering, CASE 2019
CountryCanada
CityVancouver
Period22/08/201926/08/2019
SponsorABB Robotics, University of Wisconsin–Madison, et al., IEEE, IEEE Robotics and Automation Society, National Science Foundation (NSF)
SeriesIEEE International Conference on Automation Science and Engineering
Volume2019-August
ISSN2161-8070

Fingerprint Dive into the research topics of 'Improving the Generalizability of Robot Assembly Tasks Learned from Demonstration via CNN-based Segmentation'. Together they form a unique fingerprint.

Cite this