Research Interests
- Embodied AI and Robot Learning
- World Action Models (WAM)
- Vision-Language-Action Models (VLA)
- Robot Foundation Models
- Representation Learning and 3D Vision
- Dexterous and Contact-rich Manipulation
Education
-
2025 - PresentInstitute of Automation, Chinese Academy of Sciences / Beijing Academy of Artificial Intelligence
-
2021 - 2025Sun Yat-sen University
Publications
*: joint first author; † project lead; #: corresponding author(s)
Emerging Extrinsic Dexterity in Cluttered Scenes via Dynamics-aware Policy Learning
By leveraging dynamics-aware policy learning, our approach enables robots to exploit environmental contacts and object interactions to accomplish complex manipulation.
StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision
Stereo geometric cues for improving spatial perception and viewpoint generalization in VLA models.
LDA-1B: Scaling Latent Dynamics Action Model via Universal Embodied Data Ingestion
Scaling robot foundation models with large-scale heterogeneous embodied data.