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

Download:

Current browse context:

cs.RO

Change to browse by:

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: Human-in-the-loop Robotic Grasping using BERT Scene Representation

Abstract: Current NLP techniques have been greatly applied in different domains. In this paper, we propose a human-in-the-loop framework for robotic grasping in cluttered scenes, investigating a language interface to the grasping process, which allows the user to intervene by natural language commands. This framework is constructed on a state-of-the-art rasping baseline, where we substitute a scene-graph representation with a text representation of the scene using BERT. Experiments on both simulation and physical robot show that the proposed method outperforms conventional object-agnostic and scene-graph based methods in the literature. In addition, we find that with human intervention, performance can be significantly improved.
Comments: 15 pages, 10 figures, Coling2022 Oral
Subjects: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
Cite as: arXiv:2209.14026 [cs.RO]
  (or arXiv:2209.14026v1 [cs.RO] for this version)

Submission history

From: Yaoxian Song [view email]
[v1] Wed, 28 Sep 2022 12:16:29 GMT (4515kb,D)

Link back to: arXiv, form interface, contact.