Zengyi Qin

Papers

Total Citations

113

H-Index

About

Zengyi Qin is a robotics researcher whose work bridges the critical gap between perception and control, enabling machines to interact with the world more intelligently and safely. His primary research areas include robot manipulation, representation learning, and safe control theory. Qin’s most influential contribution is the **KETO framework**, which teaches robots to learn keypoint representations for manipulating novel objects as tools. This work, which has garnered over 86 combined citations, allows a robot to understand an object’s functional geometry—like the tip of a hammer or the scoop of a spoon—without prior knowledge, dramatically improving generalization for task completion. In parallel, Qin has advanced the safety of autonomous systems through his work on **Robust Neural Lyapunov-Barrier Functions**. This model-based approach synthesizes feedback controllers that provide formal guarantees for both stability and safety, even under nonlinear and uncertain dynamics. By integrating deep learning with classical control theory, his research offers a principled path toward deploying robots in unstructured, human-centric environments. Qin’s contributions are foundational for creating robots that are not only dexterous but also provably safe.