Feel Robot Feels: Tactile Feedback Array Glove for Dexterous Manipulation

Feiyu Jia^1,2,*, Xiaojie Niu^1,*, Sizhe Yang^1,3,*, Qingwei Ben^1,3, Tao Huang^4,1,
Feng Zhao^2,†, Jingbo Wang^2,†, Jiangmiao Pang^2
1Shanghai AI Laboratory, ²University of Science and Technology of China,
³The Chinese University of Hong Kong, ⁴Shanghai Jiao Tong University
^* Equal Contributions, ^† Corresponding Author

📋 Table of Contents

• 🏠 About
• 🔥 News
• 📝 TODO List
• 🛠️ Installation
• 🕹️ Quickly Start
• 📄 License
• 👏 Acknowledgements
• 🔗 Citation

🔥 News

• [2026-04] We release hardware components, control code and hand retargeting code.
• [2026-03] We release the paper and demos of TAG.

🏠 About

This repository is an official implementation of "Feel Robot Feels: Tactile Feedback Array Glove for Dexterous Manipulation". TAG (Tactile Feedback Array Glove) presents a system-level contribution by seamlessly integrating high-accuracy motion capture with high-resolution tactile feedback into a single, practical teleoperation interface.

This repository contains three key components of TAG:

• Hardware:
  • 3D_Models: TAG .step files. You can also view them here
  • Electronic_Components: Schematics and PCB files for all TAG modules, including 21DoF_Glove_mainboard, Joint_module_MLX90393, HV507_mainboard, HV507_clip, Tactile_feedback_module, and LeapTac (including LeapTac_film and LeapTac_mainboard)
• Control:
  • Glove: Host-side program for decoding glove data(Glove_Receive.py) and mapping implementation for tactile feedback
  • STM32_Firmware: STM32 CubeIDE project files, including 21DoF_Glove, Tactile_Feedback_Module, Tactile_Feedback_Five_Finger, LeapTac_Module
• Retargeting:
  • Arm_Retargeting: Code for retargeting HTC Vive Tracker motion to G1 and UR5e
  • Hand_Retargeting: Code for retargeting the 21DoF Glove to multiple dexterous hands, including simulation and real-hardware setups
  • URDF: .urdf of the 21DoF Glove

📝 TODO List

• Release the paper with demos.
• Release the hardware components, including the 3D models, PCB files, and control code.
• Release hand retargeting code.
• Release arm retargeting code.
• Update with more detailed hardware fabrication guides and videos.

🛠️ Installation

1. Clone this repository:

git clone https://github.com/InternRobotics/TAG.git
cd TAG

2. Create a conda environment:

conda create -n tag python=3.10
conda activate tag

3. Install Required Python Packages:

pip install -r requirements.txt

4. Install pinocchio and nlopt:

conda install -c conda-forge pinocchio
conda install -c conda-forge nlopt

5. Install XHand python SDK:

You can follow the official guideline：

1. Download sdk and extract.

2. Install C/C++dependencies:

   sudo apt update && sudo apt install -y \
   cmake \
   g++ \
   libcurl4-openssl-dev \
   libssl-dev \
   nlohmann-json3-dev

3. Install xhand_controller:

   cd xhand_control_sdk_py
   pip install -r requirements.txt
   pip install xhand_controller-*-cp310-cp310-*.whl

🕹️ Quickly Start

1. Test in simulation using recorded data:

• Old_method:

  cd Retargeting/Hand_Retargeting/Old_method
  python sim_play_leaphand.py

• New_method:

  cd Retargeting/Hand_Retargeting/New_method
  python sim_play_xhand.py

2. Realtime Retargeting:

• Take XHand as an example: connect the XHand USB and TAG USB to the host, then check the /dev/ttyUSB* assignments.
• Add permissions:

  sudo chmod 666 /dev/ttyUSB*

• Run realtime retargeting:

  cd Retargeting/Hand_Retargeting
  python Realtime_Retargeting_xhand.py

If you have any questions about this repository, please feel free to drop an e-mail at feiyujia635@gmail.com.

📄 License

All code of TAG is under the CC BY-NC-SA 4.0 International License . It is strictly forbidden to use it for commercial purposes before asking our team.

👏 Acknowledgements

• HOMIE: We use HomieHardware library to build the HomieGlove.
• AnyTeleop: We use the dex-retargeting library as a visual baseline for teleoperation tasks. (This part was added in the rebuttal and will be merged into the camera-ready version later.)
• VIVE tracker: We use SteamVR_Tracking library for wrist tracking.

🔗 Citation

If you find our work helpful, please cite:

@article{jia2026feel,
  title   = {Feel Robot Feels: Tactile Feedback Array Glove for Dexterous Manipulation},
  author  = {Jia, Feiyu and Niu, Xiaojie and Yang, Sizhe and Ben, Qingwei and Huang, Tao and Zhao, Feng and Wang, Jingbo and Pang, Jiangmiao},
  journal = {arXiv preprint arXiv:2603.28542},
  year    = {2026},
}