# DOGlove: Dexterous Manipulation with a Low-Cost Open-Source Haptic Force Feedback Glove

**Authors:** Han Zhang, Songbo Hu, Zhecheng Yuan, Huazhe Xu  
**arXiv:** [2502.07730](https://arxiv.org/abs/2502.07730)  
**Date:** 2025-02-11  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2502.07730v1/

## Key points

- Dexterous hand teleoperation is crucial for robots to achieve manipulation dexterity comparable to humans
- Existing teleoperation systems are expensive and lack multi-modal sensory feedback, limiting operators' ability
- DOGlove is a cost-effective, precise, and haptic force feedback glove system designed for teleoperation and manipulation tasks
- Standout features of DOGlove include affordability, ease of assembly (under 600 USD), capturing motion across 21 degrees of freedom (DoF), multidirectional force feedback across 5-DoF, and haptic feedback at the fingertips across another 5-DoF
- DOGlove facilitates precise and immersive teleoperation of robotic hands with high success rates in executing complex manipulation tasks
- Haptic force feedback plays a critical role in enhancing task performance even without visual feedback
- Data collected from demonstrations with DOGlove can effectively train imitation learning policies, advancing machine learning algorithms in robotic applications
- Hardware and software components of DOGlove will be fully open-sourced on their website https://do-glove.github.io/
- This development paves the way for affordable yet highly functional teleoperation systems that could revolutionize how robots interact with their environments

## Summary

In the realm of robotics, dexterous hand teleoperation is crucial for enabling robots to achieve manipulation dexterity comparable to that of humans. However, existing teleoperation systems often come with a hefty price tag and lack multi-modal sensory feedback, which limits human operators' ability to accurately perceive object properties and execute complex manipulation tasks. To combat these limitations, a team of researchers comprising Han Zhang, Songbo Hu, Zhecheng Yuan, and Huazhe Xu have introduced DOGlove - a cost-effective, precise, and haptic force feedback glove system designed for teleoperation and manipulation tasks. One of the standout features of DOGlove is its affordability and ease of assembly; it can be put together in a matter of hours at a cost under 600 USD. The glove boasts a customized joint structure capable of capturing motion across 21 degrees of freedom (DoF), along with a compact cable-driven torque transmission mechanism that provides multidirectional force feedback across 5-DoF. Additionally, DOGlove incorporates a linear resonate actuator for delivering haptic feedback at the fingertips across another 5-DoF. By leveraging action and haptic force retargeting techniques, DOGlove facilitates precise and immersive teleoperation of sophisticated robotic hands. This capability has been demonstrated through high success rates in executing complex manipulation tasks involving rich contact interactions. Furthermore, the researchers conducted evaluations using scenarios devoid of visual feedback to underscore the critical role played by haptic force feedback in enhancing task performance. In an innovative move, the team utilized data collected from demonstrations with DOGlove to train imitation learning policies effectively. This highlights the potential efficacy of DOGlove not just as a standalone tool but also as an enabler for advancing machine learning algorithms in robotic applications. To promote transparency and encourage further research collaboration, the hardware and software components of DOGlove will be fully open-sourced on their website https://do-glove.github.io/. Overall, this groundbreaking development paves the way for affordable yet highly functional teleoperation systems that could revolutionize how robots interact with their environments in various industries.