JAIST Repository >
School of Information Science >
Articles >
Journal Articles >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10119/18388

Title: A Few-Shot Learning Framework for Planar Pushing of Unknown Objects
Authors: Gao, Ziyan
Elibol, Armagan
Chong, Nak Young
Keywords: Planar Pushing
Path Planning
Non-prehensile Manipulation
Data-driven Automation
Issue Date: 2022-05-21
Publisher: Springer Nature
Magazine name: Intelligent Service Robotics
Volume: 15
Start page: 335
End page: 350
DOI: 10.1007/s11370-022-00425-7
Abstract: Robot planar pushing is one of the primitive elements of non-prehensile manipulation skills, and has been widely studied as an alternative solution to complex manipulation tasks. To transfer this skill to novel objects, reasoning the pushing effect on object motion is important for selecting proper contact locations and pushing directions. However, complex contact conditions and unknown physical properties of the object cause difficulties in reasoning. In this work, firstly, we present a new large planar pushing dataset that contains a wide range of simulated objects, and a novel representation for pushing primitives for the data-driven prediction model. Secondly, we propose a computation efficient planning method that employs a heuristic to reduce the possibility of making sliding contact between the pusher and the object. The prediction model and planning method were evaluated both in simulation and real experimental settings. The results show that the prediction model purely trained using our simulation dataset is capable of predicting real object motions accurately. The push planning method effectively reduces the number of pushes required to move unknown real objects to target positions.
Rights: This is the author's version of the work. Copyright (C) 2022, Ziyan Gao, Armagan Elibol, Nak Young Chong, under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature. The version published by Springer-Verlag is available at www.springerlink.com, https://link.springer.com/article/10.1007/s11370-022-00425-7
URI: http://hdl.handle.net/10119/18388
Material Type: author
Appears in Collections:b10-1. 雑誌掲載論文 (Journal Articles)

Files in This Item:

File Description SizeFormat
N-CHONG-I-0523.pdf6147KbAdobe PDFView/Open

All items in DSpace are protected by copyright, with all rights reserved.

 


Contact : Library Information Section, JAIST (ir-sys[at]ml.jaist.ac.jp)