Kartic Subr
Papers
8
Total Citations
92
H-Index
5
About
Kartic Subr is a researcher working at the intersection of robotics, machine learning, and computer vision, with particular focus on robot learning, physical simulation, and dynamic scene understanding. His most influential contribution explores adaptive robot manipulation through fluid simulation, demonstrating how robots can learn to perform dexterous pouring tasks using fast approximate physics models — a paper that has garnered 29 citations and highlights his interest in bridging physical reasoning with autonomous behavior. Subr has made notable contributions to reward learning and inverse reinforcement learning, developing probabilistic temporal ranking methods that enable robots to infer task goals from exploratory demonstrations without predefined reward functions, with applications ranging from informative path planning to robotic ultrasound scanning. His Vid2Param research thread represents another significant strand of his work, showing how dynamical system parameters can be identified directly from video streams — enabling robots to reason about their environments in real time. His involvement in co-organizing CoRL 2017, one of robotics' premier learning-focused venues, further underscores his standing in the community. Collectively, his work advances the capacity of robots to learn, adapt, and reason in complex, uncertain real-world settings.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3The 1st Annual Conference on Robot Learning (CoRL 2017)14 citations · 2017
- 4Vid2Param: Modeling of Dynamics Parameters From Video10 citations · 2019
- 5
- 6
- 7Vid2Param: Modelling of Dynamics Parameters from Video5 citations · 2019
- 8