Haoyang Lu
Papers
3
Total Citations
8
H-Index
2
About
Haoyang Lu is pioneering the intersection of computer vision, foundation models, and robotic manipulation, with a focus on enabling robots to learn complex skills from abundant, low-cost data sources. His central research thrust is overcoming the critical bottleneck of expensive, action-labeled robot data by leveraging human video demonstrations. Lu’s major contributions include developing novel frameworks like GraphMimic and FMimic, which harness graph-based generative modeling and vision-language models (VLMs) to translate passive video observations into fine-grained, executable robot policies. This work directly addresses the challenge of data scarcity in robotics, opening pathways for scalable skill acquisition. Additionally, his research on high-precision object pose estimation integrates visual and tactile sensing to achieve robust performance in dynamic, contact-rich tasks such as assembly and grasping. Though his most-cited papers are recent (2025), they have already garnered early attention (4 and 3 citations respectively), signaling strong and growing impact in the field. Lu’s work is notable for its practical, data-driven approach to bridging the simulation-to-reality gap, making him a rising figure in the quest for generalist robots that learn from the world as humans do.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3