Shuaijun Wang

Beijing, China

📧 wukongwoong@gmail.com

“Noting new in human intelligence compared to AI.”

Hello, I am Shuaijun Wang, currently serving as Technical Lead of Dexterous Manipulation in the Robotics Department at Xiaomi Technology. I received my Ph.D. degree from the Robotics Institute, National Key Laboratory of Harbin Institute of Technology, under the supervision of Professor Lining Sun. I obtained my M.S. degree from Harbin Institute of Technology(Shenzhen) under the supervision of Professor Xin Wang, and my B.S. degree from Northeast Forestry University as an Outstanding Graduate.

Between 2019 and 2021, I conducted research as a visiting scholar at the School of Informatics, Robotics Center of the University of Edinburgh, working with Professor Zhibin(Alex) Li. From March 2022 to December 2022, I collaborated remotely with Professor Qiang Li at the Robotics Center of the University of Bielefeld in Germany. In 2022-2023, I worked as a research intern at Tencent Robotics X Laboratory.

My research focuses on universal and dexterous grasping and manipulation of humanoid robots, as well as embodied AI.

Research Interests

Dexterous Manipulation and Grasping
Tactile Perception and Feedback
Reinforcement Learning and Imitation Learning
Full-self Manipulation(FSM)
Embodied Intelligence

news

Mar 29, 2026	Our work TacRefineNet is featured in Xiaomi Group’s FY2025 annual results presentation materials (PDF).
Oct 15, 2025	Our paper “HD²Handover: Human-like Dynamic Dexterous Bimanual Handover using RGB-only Perception” has been accepted to the IROS 2025 Workshop on Bimanual Manipulation: Advancing Human-Humanoid Interaction and Collaboration.
Jul 01, 2024	Joined Xiaomi Robotics Department as a Senior Algorithm Engineer, building from 0-1 and serving as Technical Lead of the Dexterous Manipulation Team! 🎉
May 01, 2024	Successfully earned Ph.D. in Robotic Learning from Harbin Institute of Technology! 🎓
Dec 01, 2023	Invited talk at Shenzhen Technology University on “Embodied Intelligence Methods for General-Purpose Robotic Grasping and Manipulation”.(Received Assistant Professor invitation)
Jun 01, 2023	Completed a research internship at Tencent Robotics X Lab (Dec. 2022 – Jun. 2023), where an existing fingertip tactile-based in-hand manipulation algorithm was reproduced and successfully transferred to Tencent’s next-generation dexterous hand platform, receiving broad media coverage.
Feb 16, 2022	Paper “Learning Adaptive Grasping From Human Demonstrations” published in IEEE/ASME Transactions on Mechatronics (SCI Q1 Top journal in Robotics)! 🏆
Sep 15, 2021	I completed a two-year CSC joint Ph.D. training program at the School of Informatics, University of Edinburgh (Robotics Centre) from September 2019 to September 2021.

selected publications

arXiv
TacRefineNet: Tactile-Only Grasp Refinement Between Arbitrary In-Hand Object Poses(IEEE RAL under review)

Shuaijun Wang^*, Haoran Zhou, Diyun Xiang, and 1 more author

arXiv preprint arXiv:2509.25746, 2025

Abs Bib HTML Website

This paper presents TacRefineNet, a novel approach for tactile-only grasp refinement that enables precise manipulation between arbitrary in-hand object poses using only tactile feedback.
@article{wang2025tacrefinenet, title = {TacRefineNet: Tactile-Only Grasp Refinement Between Arbitrary In-Hand Object Poses(IEEE RAL under review)}, author = {Wang, Shuaijun and Zhou, Haoran and Xiang, Diyun and You, Yangwei}, journal = {arXiv preprint arXiv:2509.25746}, year = {2025}, }
arXiv
DyDexHandover: Human-like Bimanual Dynamic Dexterous Handover using RGB-only Perception(ICRA2026 under review)

Haoran Zhou, Yangwei You, and Shuaijun Wang^*

arXiv preprint arXiv:2509.17350, 2025

Abs Bib HTML Website

This work presents DyDexHandover, a novel approach for human-like bimanual dynamic dexterous handover using only RGB perception, enabling natural and efficient object transfer between robotic hands.
@article{zhou2025dydexhandover, title = {DyDexHandover: Human-like Bimanual Dynamic Dexterous Handover using RGB-only Perception(ICRA2026 under review)}, author = {Zhou, Haoran and You, Yangwei and Wang, Shuaijun}, journal = {arXiv preprint arXiv:2509.17350}, year = {2025}, }

IJHR

Learning an Image-Based Visual Servoing Controller for Object Grasping

Shuaijun Wang, Lining Sun, Mantian Li, and 4 more authors

International Journal of Humanoid Robotics, 2024

SCI

Abs DOI Bib HTML Video

@article{wang2024learning,
  title = {Learning an Image-Based Visual Servoing Controller for Object Grasping},
  author = {Wang, Shuaijun and Sun, Lining and Li, Mantian and Wang, Pengfei and Zha, Fusheng and Guo, Wei and Li, Qiang},
  journal = {International Journal of Humanoid Robotics},
  volume = {21},
  number = {05},
  pages = {2350033},
  year = {2024},
  publisher = {World Scientific},
  doi = {10.1142/S0219843623500337},
  note = {SCI}
}

Front. Neurorobot.
Learning Adaptive Reaching and Pushing Skills Using Contact Information

Shuaijun Wang, Lining Sun, Fusheng Zha, and 2 more authors

Frontiers in Neurorobotics, 2023

SCI Q2

Abs DOI Bib HTML Video

We develop a reinforcement learning-based pushing strategy using Soft Actor-Critic (SAC) with explicit contact modeling. The approach demonstrates improved generalization and training efficiency for contact-aware object manipulation tasks.
@article{wang2023learning, title = {Learning Adaptive Reaching and Pushing Skills Using Contact Information}, author = {Wang, Shuaijun and Sun, Lining and Zha, Fusheng and Guo, Wei and Wang, Pengfei}, journal = {Frontiers in Neurorobotics}, volume = {17}, pages = {112976}, year = {2023}, publisher = {Frontiers Media SA}, doi = {10.3389/fnbot.2023.1129076}, note = {SCI Q2} }
TMECH
Learning Adaptive Grasping From Human Demonstrations

Shuaijun Wang, Wenbin Hu, Lining Sun, and 2 more authors

IEEE/ASME Transactions on Mechatronics, 2022

SCI Q1

Abs DOI Bib HTML Video

This paper presents a learning-based approach to enable adaptive grasping from human demonstrations. We collect human teleoperation data and use supervised learning to train neural networks that can handle object variability in size and stiffness. The proposed method demonstrates robust performance across different object categories.
@article{wang2022learning, title = {Learning Adaptive Grasping From Human Demonstrations}, author = {Wang, Shuaijun and Hu, Wenbin and Sun, Lining and Wang, Xin and Li, Zhibin}, journal = {IEEE/ASME Transactions on Mechatronics}, volume = {27}, number = {6}, pages = {5148--5159}, year = {2022}, publisher = ieee, doi = {10.1109/TMECH.2022.3175885}, note = {SCI Q1} }