|
Zhenyu Wei | ๅซๆฏๅฎ I am a first-year CS Ph.D. student at University of North Carolina at Chapel Hill, advised by Prof. Mingyu Ding. I received my Bachelor's degree in Computer Science at at Shanghai Jiao Tong University. During my undergraduate years, I was fortunate to be advised by Prof. Lin Shao at National University of Singapore and Prof. Cewu Lu at Shanghai Jiao Tong University. Email / CV / Github / Google Scholar / WeChat
×
|
|
News
|
ResearchMy research interests lie in ๐คdexterous manipulation. My long-term goal is to make robots an essential part of daily life, reshaping how we live. I'm open to collaborations on robotics related projects, feel free to contact me!๐ค Papers sorted by recency. Representative papers are highlighted. |
|
RL-100: Performant Robotic Manipulation with Real-World Reinforcement Learning
*, *, *, *, , , , , Website / arXiv
TL;DR:
RLโ100 introduces a three-stage real-world learning pipeline combining imitation learning, offline RL, and online RL,
leveraging learned visuomotor policies to achieve 100% task success across seven diverse manipulation tasks, while
enabling strong zero- and few-shot generalization to new dynamics and perturbations.
|
|
TelePreview: A User-Friendly Teleoperation System with Virtual Arm Assistance for Enhanced Effectiveness
, , , , , , , Website / arXiv
TL;DR:
Implement a low-cost teleoperation system utilizing data gloves and IMU sensors, paired with an assistant
module that improves data collection process by visualizing future robot operations through visual previews.
|
|
MetaFold: Language-Guided Multi-Category Garment Folding Framework via Trajectory Generation and Foundation Model
*, *, , , , , , , , , , Website / arXiv
IROS 2025
International Conference on Intelligent Robots and Systems
Oral Presentation
TL;DR:
MetaFold employs language-guided point cloud trajectories for task planning and a foundation model for
action prediction, enabling better generalization across garments and instructions.
|
|
$\mathcal{D(R,O)}$ Grasp: A Unified Representation of Robot and Object Interaction for Cross-Embodiment Dexterous Grasping
*, *, , , , , , Website / arXiv / Code / Media (ๆบๅจไนๅฟ)
ICRA 2025
International Conference on Robotics and Automation
ICRA 2025 Best Paper Award on Robot Manipulation and Locomotion
Best Robotics Paper Award, CoRL 2024 @ MAPoDeL
×
×
TL;DR:
Introduce $\mathcal{D(R,O)}$, a novel interaction-centric representation for dexterous grasping tasks that
goes beyond traditional robot-centric and object-centric approaches, enabling robust generalization across
diverse robotic hands and objects.
|
|
Auto-Pairing Positives through Implicit Relation Circulation for Discriminative Self-Learning
, , , IEEE Xplore / Code
TPAMI 2025
IEEE Transactions on Pattern Analysis and Machine Intelligence
TL;DR:
Introduce Implicit Relation Circulation (IRC), a self-supervised learning framework that discovers positive pairs
automatically by performing random walks across feature groups and enforcing cycle consistency with a simpler task,
without requiring explicit matching.
|
Selected Honors
|
Experience |
|
University of North Carolina at Chapel Hill, USA 2025.08 - PresentPh.D. in Computer ScienceResearch Advisor: Prof. Mingyu Ding |
|
Shanghai Qi Zhi Institute, China 2025.02 - 2025.07Research InternResearch Advisor: Prof. Huazhe Xu |
|
NUS LinS Lab, Singapore 2024.06 - 2025.02Research AssistantResearch Advisor: Prof. Lin Shao |
|
SJTU Machine Vision and Intelligence Group (MVIG), China 2022.10 - 2024.05Research InternResearch Advisor: Prof. Cewu Lu |
|
|
Shanghai Jiao Tong University, China 2021.09 - 2025.06B.Eng. in Computer Science (Zhiyuan Honor Degree)GPA: 4.0/4.3 |
|
Thank you for visiting ๐ Feel free to contact me if you have any questions.
|