Carlos Celemin
Papers
12
Total Citations
201
H-Index
8
About
Carlos Celemin is a prominent robotics researcher whose work sits at the intersection of interactive machine learning, imitation learning, and reinforcement learning, with a particular focus on enabling non-expert humans to intuitively train and adapt robotic systems. His most influential contribution, a comprehensive survey on Interactive Imitation Learning (IIL) in robotics (2022, 53 citations), has helped establish IIL as a distinct and rapidly growing subfield, synthesizing how intermittent human feedback can drive online improvement of robot behavior. Celemin has made significant strides in developing hybrid frameworks that combine human corrective advice with policy search reinforcement learning, demonstrating faster and more reliable learning in real-world robotic motor skills — work recognized through strong citation counts across multiple publications from 2018 and 2019. His COACH framework introduced an elegant mechanism for shaping continuous action policies through binary human signals, broadening accessibility for non-expert users. Later contributions tackled nuanced challenges including temporal feature learning, high-dimensional state spaces, ambiguity resolution, and robustness to imperfect human teachers. With a cumulative body of work exceeding 190 citations, Celemin's research meaningfully advances the vision of robots that can be flexibly programmed and adapted by everyday users.
Research Focus
Key Achievements
Top Papers
- 1Interactive Imitation Learning in Robotics: A Survey53 citations · 2022
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9Learning Interactively to Resolve Ambiguity in Reference Frame Selection6 citations · 2020
- 10Interactive Imitation Learning in Robotics: A Survey6 citations · 2022