Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Impact of Object Weight in Handovers: Inspiring Robotic Grip Release and Motion from Human Handovers

Parag Khanna, Mårten Björkman, Christian Smith

TL;DR

This work investigates how object weight shapes human and robotic handovers, combining human experiments with two weight-diverse datasets (Handovers@RPL-2.0 and YCB Handovers) to quantify effects on grip release, transfer timing, and motion. It develops weight-aware adaptive strategies for both robotic motion and grip release, including a data-driven LSTM-based grip-release predictor and a VAE-LSTM-based adaptive release controller, evaluated in robot-to-human handovers. Key findings show that heavier objects increase transfer and grip-release times while reducing motion speeds, and that weight-aware, human-inspired strategies improve perceived naturalness, weight anticipation, and user confidence, especially for moderate-to-heavy objects. The work contributes weight-diverse datasets, weight-informed motion adaptation, and data-driven grip-release methods to advance natural, safe, and efficient human-robot handovers in real-world settings.

Abstract

This work explores the effect of object weight on human motion and grip release during handovers to enhance the naturalness, safety, and efficiency of robot-human interactions. We introduce adaptive robotic strategies based on the analysis of human handover behavior with varying object weights. The key contributions of this work includes the development of an adaptive grip-release strategy for robots, a detailed analysis of how object weight influences human motion to guide robotic motion adaptations, and the creation of handover-datasets incorporating various object weights, including the YCB handover dataset. By aligning robotic grip release and motion with human behavior, this work aims to improve robot-human handovers for different weighted objects. We also evaluate these human-inspired adaptive robotic strategies in robot-to-human handovers to assess their effectiveness and performance and demonstrate that they outperform the baseline approaches in terms of naturalness, efficiency, and user perception.

Impact of Object Weight in Handovers: Inspiring Robotic Grip Release and Motion from Human Handovers

TL;DR

This work investigates how object weight shapes human and robotic handovers, combining human experiments with two weight-diverse datasets (Handovers@RPL-2.0 and YCB Handovers) to quantify effects on grip release, transfer timing, and motion. It develops weight-aware adaptive strategies for both robotic motion and grip release, including a data-driven LSTM-based grip-release predictor and a VAE-LSTM-based adaptive release controller, evaluated in robot-to-human handovers. Key findings show that heavier objects increase transfer and grip-release times while reducing motion speeds, and that weight-aware, human-inspired strategies improve perceived naturalness, weight anticipation, and user confidence, especially for moderate-to-heavy objects. The work contributes weight-diverse datasets, weight-informed motion adaptation, and data-driven grip-release methods to advance natural, safe, and efficient human-robot handovers in real-world settings.

Abstract

This work explores the effect of object weight on human motion and grip release during handovers to enhance the naturalness, safety, and efficiency of robot-human interactions. We introduce adaptive robotic strategies based on the analysis of human handover behavior with varying object weights. The key contributions of this work includes the development of an adaptive grip-release strategy for robots, a detailed analysis of how object weight influences human motion to guide robotic motion adaptations, and the creation of handover-datasets incorporating various object weights, including the YCB handover dataset. By aligning robotic grip release and motion with human behavior, this work aims to improve robot-human handovers for different weighted objects. We also evaluate these human-inspired adaptive robotic strategies in robot-to-human handovers to assess their effectiveness and performance and demonstrate that they outperform the baseline approaches in terms of naturalness, efficiency, and user perception.

Paper Structure

This paper contains 42 sections, 6 equations, 25 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Datasets
  4. Human Motion and Joint Coordination
  5. Forces in Handovers and Grip Release
  6. Human Handovers Study and Datasets
  7. Experimental Setup
  8. Handovers with sensor-embedded object
  9. Sensor-Embedded Object
  10. Experimental Settings
  11. Participants
  12. Handovers with generic Objects: YCB Handovers Dataset
  13. YCB dataset objects
  14. Experimental Settings
  15. Participants
  16. ...and 27 more sections

Figures (25)

  • Figure 1: An overview of our approach: We investigate the impact of object weight on human motion and grip-release dynamics in human-human handovers. Based on these insights, we propose a data-driven strategy for adaptive grip release and evaluate its effectiveness in robot-human handovers. Additionally, we develop and assess weight-adaptive motion strategies for robots during handovers. This comprehensive approach aims to enhance the naturalness and efficiency of robot-human object handovers across various weight categories.
  • Figure 2: 3D printed sensor embedded Baton
  • Figure 3: (a) Table (0.8 m wide) setup for human-human handovers (b) Precision grasp on baton with lead weights added
  • Figure 4: Grip force ($F^g$) variation in a particular handover.
  • Figure 5: Mean transfer time
  • ...and 20 more figures