Sanghoon Lee
Amazon (United States), Hanyang University, Yonsei University
Papers
10
Total Citations
140
H-Index
5
About
Sanghoon Lee is a versatile robotics and artificial intelligence researcher whose work spans differentiable physics simulation, reinforcement learning, computer vision, and human-robot interaction. Perhaps his most significant contribution is the development of **Nimble**, a fast and feature-complete differentiable physics engine for articulated rigid body simulation with contact constraints, introduced in 2021 and accumulating over 60 citations. Nimble supports Lagrangian dynamics and hard contact constraints, offering capabilities previously unavailable in differentiable simulation frameworks — making it a valuable tool for the robotics and machine learning communities. Earlier in his career, Lee made foundational contributions to behavior-based robot control and reinforcement learning, proposing novel action-selection mechanisms and stochastic shortest path Q-learning (SSPQL) approaches that addressed sequential decision-making and delayed-reward challenges in mobile robotics. His work on wearable knee-assistive exoskeletons further demonstrated his interest in human locomotion and assistive technology. More recently, he has contributed to computer vision problems relevant to robotics, including facial feature tracking using adaptive particle filters and deep learning-based chessboard corner detection for camera calibration. With citations spanning nearly two decades of research, Lee's work reflects a sustained commitment to bridging intelligent learning systems with physical robotic applications.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4SSPQL: Stochastic shortest path-based Q-learning12 citations · 2011
- 5
- 6
- 7
- 8Deep Chessboard Corner Detection Using Multi-task Learning4 citations · 2021
- 9
- 10