Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Getting a Grip: Learning Precise Robot-To-Human Handovers with Hand Pose Conditioning

Ertürk, Esranur LU ; Stenmark, Maj LU orcid and Topp, Elin Anna LU orcid (2026) 2026 IEEE 35th International Conference on Robot and Human Interactive Communication (RO-MAN)
Abstract (Swedish)
Robot-to-human handovers are fundamental for collaborative robots, particularly in surgical contexts where instruments must be delivered accurately without regrasping.

We investigate the role of hand representation in learning precise robot-to-human handover behaviors in complex contexts such as surgery. Using a dataset of 180 teleoperated demonstrations, an image-based Action Chunking with Transformers (ACT) policy is compared to a variant conditioned on explicit 3D hand pose information extracted from visual observations.

Our findings highlight the importance of representation in close human-robot interaction tasks and suggest that explicit hand pose information is critical for learning precise and usable robot-to-human... (More)
Robot-to-human handovers are fundamental for collaborative robots, particularly in surgical contexts where instruments must be delivered accurately without regrasping.

We investigate the role of hand representation in learning precise robot-to-human handover behaviors in complex contexts such as surgery. Using a dataset of 180 teleoperated demonstrations, an image-based Action Chunking with Transformers (ACT) policy is compared to a variant conditioned on explicit 3D hand pose information extracted from visual observations.

Our findings highlight the importance of representation in close human-robot interaction tasks and suggest that explicit hand pose information is critical for learning precise and usable robot-to-human handovers in surgical settings. (Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Contribution to conference
publication status
published
subject
conference name
2026 IEEE 35th International Conference on Robot and Human Interactive Communication (RO-MAN)
conference location
Fukuoka, Japan
conference dates
2026-08-24 - 2026-08-28
language
English
LU publication?
yes
id
4b73e3a1-8996-44dd-b1f0-790b256c0d77
alternative location
https://ras.papercept.net/conferences/conferences/ROMAN26/program/ROMAN26_ContentListWeb_3.html
date added to LUP
2026-10-07 18:06:27
date last changed
2026-10-08 07:10:37
@misc{4b73e3a1-8996-44dd-b1f0-790b256c0d77,
  abstract     = {{Robot-to-human handovers are fundamental for collaborative robots, particularly in surgical contexts where instruments must be delivered accurately without regrasping.<br/><br/>We investigate the role of hand representation in learning precise robot-to-human handover behaviors in complex contexts such as surgery. Using a dataset of 180 teleoperated demonstrations, an image-based Action Chunking with Transformers (ACT) policy is compared to a variant conditioned on explicit 3D hand pose information extracted from visual observations.<br/><br/>Our findings highlight the importance of representation in close human-robot interaction tasks and suggest that explicit hand pose information is critical for learning precise and usable robot-to-human handovers in surgical settings.}},
  author       = {{Ertürk, Esranur and Stenmark, Maj and Topp, Elin Anna}},
  language     = {{eng}},
  title        = {{Getting a Grip: Learning Precise Robot-To-Human Handovers with Hand Pose Conditioning}},
  url          = {{https://ras.papercept.net/conferences/conferences/ROMAN26/program/ROMAN26_ContentListWeb_3.html}},
  year         = {{2026}},
}