Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Collaborative Object Handover in a Robot Crafting Assistant

Leimin Tian, Shiyu Xu, Kerry He, Rachel Love, Akansel Cosgun, Dana Kulic

TL;DR

This work tackles the challenge of safe, context-aware object handovers in human-robot collaboration by learning an autonomous handover policy from naturalistic crafting task data (FACT HRC). The authors segment handovers into action primitives and train CNN-1D classifiers to predict timing and object transfer location, evaluating the policy in a user study against teleoperated handovers. Results show the autonomous policy can achieve collaborative handovers but reveals notable gaps in efficiency and user perception compared with human teleoperation, highlighting the need for personalization and transparency. The study advances HRC by linking adaptive handover timing and localization to human cues within a rich, task-oriented context and provides open-source data and code for further development.

Abstract

Robots are increasingly working alongside people, delivering food to patrons in restaurants or helping workers on assembly lines. These scenarios often involve object handovers between the person and the robot. To achieve safe and efficient human-robot collaboration (HRC), it is important to incorporate human context in a robot's handover strategies. We develop a collaborative handover model trained on human teleoperation data collected in a naturalistic crafting task. To evaluate its performance, we conduct cross-validation experiments on the training dataset as well as a user study in the same HRC crafting task. The handover episodes and user perceptions of the autonomous handover policy were compared with those of the human teleoperated handovers. While the cross-validation experiment and user study indicate that the autonomous policy successfully achieved collaborative handovers, the comparison with human teleoperation revealed avenues for further improvements.

Collaborative Object Handover in a Robot Crafting Assistant

TL;DR

This work tackles the challenge of safe, context-aware object handovers in human-robot collaboration by learning an autonomous handover policy from naturalistic crafting task data (FACT HRC). The authors segment handovers into action primitives and train CNN-1D classifiers to predict timing and object transfer location, evaluating the policy in a user study against teleoperated handovers. Results show the autonomous policy can achieve collaborative handovers but reveals notable gaps in efficiency and user perception compared with human teleoperation, highlighting the need for personalization and transparency. The study advances HRC by linking adaptive handover timing and localization to human cues within a rich, task-oriented context and provides open-source data and code for further development.

Abstract

Robots are increasingly working alongside people, delivering food to patrons in restaurants or helping workers on assembly lines. These scenarios often involve object handovers between the person and the robot. To achieve safe and efficient human-robot collaboration (HRC), it is important to incorporate human context in a robot's handover strategies. We develop a collaborative handover model trained on human teleoperation data collected in a naturalistic crafting task. To evaluate its performance, we conduct cross-validation experiments on the training dataset as well as a user study in the same HRC crafting task. The handover episodes and user perceptions of the autonomous handover policy were compared with those of the human teleoperated handovers. While the cross-validation experiment and user study indicate that the autonomous policy successfully achieved collaborative handovers, the comparison with human teleoperation revealed avenues for further improvements.

Paper Structure

This paper contains 17 sections, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Collaborative handover model
  5. User study evaluation
  6. Machine Learning Experiments
  7. FACT HRC dataset evaluation
  8. Validating action primitives
  9. User Study Evaluation Results
  10. Handover episodes
  11. Participant behaviour and perception
  12. Teleoperated vs. Auto Handover
  13. Handover episodes
  14. Subjective perception
  15. Personalisation in teleoperated sessions
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Our autonomous handover policy performs temporal adaptations at episode start, object transfer start, and object transfer completion. Spatial adaptation of object transfer position (left-hand side of the participant, middle of the work space, right-hand side of the participant) is performed at episode and object transfer starts.
  • Figure 2: Auto vs. teleoperated handovers in terms of OTP (a), Type (b), Quality (c) and Timing (d). More teleoperated handovers adopted left or right OTP, performed bidirectional handovers, were rated as good or neutral, with longer episodes containing longer pauses at the storage space.
  • Figure 3: Handover location (a) and timing (b) choices by the operator across participants in the teleoperated HRC crafting sessions. The operator showed personalisation in their handover strategy as mixed OTPs and more diverse pauses were used in later episodes across participants.