Papers
1
Total Citations
9
H-Index
1
About
Hanbit Oh is a researcher specializing in **imitation learning, reinforcement learning, and robust policy development** for real-world robotic and autonomous systems. His most recognized work, "Disturbance-injected Robust Imitation Learning with Task Achievement" (2022), addresses a fundamental challenge in the field: the assumption that expert demonstrations are both optimal and sufficiently diverse. By introducing disturbance injection techniques, Oh's approach enables learned policies to recover from perturbations and generalize beyond the narrow distribution of training demonstrations — a critical advancement for deploying imitation learning in practical, unpredictable environments. This work has garnered 9 citations, reflecting growing interest in bridging the gap between idealized demonstration-based learning and the messy realities of real-world deployment. Oh's research challenges conventional assumptions in the imitation learning pipeline, particularly regarding demonstration quality and policy stabilization, pushing the community toward more resilient and adaptive learning frameworks. His contributions are particularly valuable for students and researchers grappling with how to make learned behaviors robust when perfect expert data is unavailable — one of the most pressing open problems in modern machine learning and robotics.
Research Focus
Key Achievements
Top Papers
- 1Disturbance-injected Robust Imitation Learning with Task Achievement9 citations · 2022