We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.RO

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Robotics

Title: Learning to Place Objects onto Flat Surfaces in Upright Orientations

Abstract: We study the problem of placing a grasped object on an empty flat surface in an upright orientation, such as placing a cup on its bottom rather than on its side. We aim to find the required object rotation such that when the gripper is opened after the object makes contact with the surface, the object would be stably placed in the upright orientation. We iteratively use two neural networks. At every iteration, we use a convolutional neural network to estimate the required object rotation, which is executed by the robot, and then a separate convolutional neural network to estimate the quality of a placement in its current orientation. Our approach places previously unseen objects in upright orientations with a success rate of 98.1% in free space and 90.3% with a simulated robotic arm, using a dataset of 50 everyday objects in simulation experiments. Real-world experiments were performed, which achieved an 88.0% success rate, which serves as a proof-of-concept for direct sim-to-real transfer.
Comments: Published in RA-L and ICRA
Subjects: Robotics (cs.RO)
DOI: 10.1109/LRA.2021.3068122
Cite as: arXiv:2004.00249 [cs.RO]
  (or arXiv:2004.00249v3 [cs.RO] for this version)

Submission history

From: Akansel Cosgun [view email]
[v1] Wed, 1 Apr 2020 06:31:18 GMT (1135kb,D)
[v2] Wed, 28 Oct 2020 04:01:08 GMT (2516kb,D)
[v3] Mon, 19 Apr 2021 17:52:13 GMT (7939kb,D)

Link back to: arXiv, form interface, contact.