Minghan Zhu
Papers
3
Total Citations
28
H-Index
3
About
Minghan Zhu is a robotics researcher whose work sits at the intersection of geometry, learning, and perception, with a focus on building systems that respect the underlying symmetries of the physical world. His research centers on three key areas: symmetry-preserving robot perception and control, equivariant learning for point cloud registration, and open-vocabulary semantic mapping. Zhu’s major contributions include pioneering the use of SE(3)-equivariant transformers for low-overlap point cloud registration—a notoriously difficult problem in robotics—achieving robust performance even under large initial pose errors. His 2022 article on progress in symmetry-preserving methods (13 citations) provides a comprehensive framework for integrating geometric structure into sensor registration, state estimation, and control. More recently, his 2025 work on LatentBKI introduces a novel probabilistic mapping algorithm that fuses vision-language models with continuous mapping, enabling robots to build open-dictionary semantic maps with quantifiable uncertainty—a significant leap beyond fixed-category approaches. With each publication pushing the boundaries of how robots perceive and interact with their environments, Zhu is establishing himself as a rising leader in geometrically principled robotic intelligence.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3