Vincent Kee

Draper Laboratory, Massachusetts Institute of Technology

Papers

Total Citations

211

H-Index

About

Vincent Kee is a roboticist whose work bridges perception and reconfiguration, advancing how robots perceive and adapt to complex environments. His key research areas include semantic scene understanding, pose estimation, and modular robotics, with a focus on enabling autonomous manipulation in unstructured settings. Kee’s most impactful contribution is **SegICP**, an integrated deep learning and pose estimation system that significantly improves robotic perception speed and robustness in realistic scenarios—a critical need highlighted by manipulation competitions. This work has garnered **151 citations**, underscoring its influence in the field. He also pioneered the concept of **nested reconfiguration** in modular robotics, introducing the Hinged-Tetro module and a theoretical framework that distinguishes intra-, inter-, and nested reconfigurability. These contributions, though with fewer citations (28 and 27), represent foundational steps in a novel design paradigm. More recently, his **SegICP-DSR** extends this work to dense semantic scene reconstruction, achieving millimeter-level pose accuracy. Kee’s research demonstrates a rare ability to integrate perception and hardware design, offering practical solutions for robots operating in cluttered, real-world environments.