Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ACE-F: A Cross Embodiment Foldable System with Force Feedback for Dexterous Teleoperation

Rui Yan, Jiajian Fu, Shiqi Yang, Lars Paulsen, Xuxin Cheng, Xiaolong Wang

TL;DR

ACE-F tackles the challenges of force feedback, cross-embodiment generalization, and portability in teleoperation by introducing a sensorless, virtual force feedback system built on a foldable 3-DoF leader arm, glove-based orientation capture, and augmented inverse kinematics. The approach decouples position and orientation, enabling universal end-effector retargeting across diverse robots while rendering contact cues through end-effector deviations and adaptive impedance. Extensive simulations and real-world experiments demonstrate that ACE-F improves task speed, reliability, and data quality for imitation learning, often outperforming joint-copy baselines. The work offers practical, low-cost, modular solutions with broad applicability to dexterous, contact-rich manipulation and data collection workflows.

Abstract

Teleoperation systems are essential for efficiently collecting diverse and high-quality robot demonstration data, especially for complex, contact-rich tasks. However, current teleoperation platforms typically lack integrated force feedback, cross-embodiment generalization, and portable, user-friendly designs, limiting their practical deployment. To address these limitations, we introduce ACE-F, a cross embodiment foldable teleoperation system with integrated force feedback. Our approach leverages inverse kinematics (IK) combined with a carefully designed human-robot interface (HRI), enabling users to capture precise and high-quality demonstrations effortlessly. We further propose a generalized soft-controller pipeline integrating PD control and inverse dynamics to ensure robot safety and precise motion control across diverse robotic embodiments. Critically, to achieve cross-embodiment generalization of force feedback without additional sensors, we innovatively interpret end-effector positional deviations as virtual force signals, which enhance data collection and enable applications in imitation learning. Extensive teleoperation experiments confirm that ACE-F significantly simplifies the control of various robot embodiments, making dexterous manipulation tasks as intuitive as operating a computer mouse. The system is open-sourced at: https://acefoldable.github.io/

ACE-F: A Cross Embodiment Foldable System with Force Feedback for Dexterous Teleoperation

TL;DR

ACE-F tackles the challenges of force feedback, cross-embodiment generalization, and portability in teleoperation by introducing a sensorless, virtual force feedback system built on a foldable 3-DoF leader arm, glove-based orientation capture, and augmented inverse kinematics. The approach decouples position and orientation, enabling universal end-effector retargeting across diverse robots while rendering contact cues through end-effector deviations and adaptive impedance. Extensive simulations and real-world experiments demonstrate that ACE-F improves task speed, reliability, and data quality for imitation learning, often outperforming joint-copy baselines. The work offers practical, low-cost, modular solutions with broad applicability to dexterous, contact-rich manipulation and data collection workflows.

Abstract

Teleoperation systems are essential for efficiently collecting diverse and high-quality robot demonstration data, especially for complex, contact-rich tasks. However, current teleoperation platforms typically lack integrated force feedback, cross-embodiment generalization, and portable, user-friendly designs, limiting their practical deployment. To address these limitations, we introduce ACE-F, a cross embodiment foldable teleoperation system with integrated force feedback. Our approach leverages inverse kinematics (IK) combined with a carefully designed human-robot interface (HRI), enabling users to capture precise and high-quality demonstrations effortlessly. We further propose a generalized soft-controller pipeline integrating PD control and inverse dynamics to ensure robot safety and precise motion control across diverse robotic embodiments. Critically, to achieve cross-embodiment generalization of force feedback without additional sensors, we innovatively interpret end-effector positional deviations as virtual force signals, which enhance data collection and enable applications in imitation learning. Extensive teleoperation experiments confirm that ACE-F significantly simplifies the control of various robot embodiments, making dexterous manipulation tasks as intuitive as operating a computer mouse. The system is open-sourced at: https://acefoldable.github.io/

Paper Structure

This paper contains 38 sections, 2 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Hardware Design.
  4. Augmented Inverse Kinematics Solver.
  5. End-Effector Control and Feedback.
  6. Force Feedback Calculation.
  7. Experiments
  8. Experiment Design.
  9. Ablation Study
  10. Baseline Simulation Experiments
  11. Simulated Box Stacking
  12. Simulated Box Dragging
  13. Simulated Table Mopping
  14. Baseline Real-World Experiments
  15. Real Can Stacking
  16. ...and 23 more sections

Figures (4)

  • Figure 1: Overview of the ACE-F system. Left: Annotated view of the ACE-F arm showing the base joint (1 DoF), perpendicular elbow joints (2 DoF), and the spherical joint for interchangeable end-effectors. Right: Several representative end-effector configurations are enabled by the spherical joint.
  • Figure 2: ACE-F's scaled virtual force-feedback control system.
  • Figure 3: Overview of the six tasks in the evaluation suite: real-world mopping, real-world can stacking, real-world blind can insertion, simulated table mopping (left-right and forward-backward), simulated box stacking, and simulated box dragging.
  • Figure 4: Overview of all four imitation learning policies: simulated lifting, simulated stacking, simulated wiping, and real-world can sorting.