Keith Grochow

Papers

Total Citations

306

H-Index

About

Keith Grochow is a pioneering researcher in the intersection of machine learning and robotics, best known for his foundational work in robotic imitation learning and latent variable modeling. His key research areas include robot learning from demonstration, human motion capture analysis, and the development of shared latent structure algorithms for cross-modal data. Grochow's most influential contribution is his 2005 paper, "Learning Shared Latent Structure for Image Synthesis and Robotic Imitation," which has garnered 183 citations. In this work, he introduced a Gaussian process regression framework that learns a common hidden structure linking heterogeneous observation spaces—a breakthrough that enabled robots to synthesize and imitate human-like movements more effectively. His 2006 follow-up, with 71 citations, further solidified imitation learning as a regression problem, while his 2007 paper provided the first demonstration of a humanoid robot learning to walk by directly imitating human gait from motion capture data, without any prior dynamics model. This achievement, with 44 citations, marked a significant step toward flexible, data-driven robotic systems. Grochow's work has profoundly influenced how robots acquire complex motor skills through observation, making him a key figure in modern robotics and machine learning.