My research goal is to develop learning-based control algorithms that enable robust and agile locomotion and interactive behaviors within human environments.
For this, I am currently working on research in legged robot locomotion, employing an approach that integrates classical control and learning-based methods.
I believe this approach can complement the drawbacks of each method by reducing the burden of reward engineering and improving the performance of the model-based method.
Furthermore, I am exploring the possibilities of human motion-driven control by establishing correspondence between humans and robots.
We propose a learning framework that can bridge between model-based and learning-based approaches for legged robot control by imitating expert model predictive control (MPC) and fine-tuning the pre-trained policy with reinforcement learning.
Ongoing Projects
Humanoid locomotion with learning-based method leveraging model-based control
Training the locomotion of full-sized humanoids is a challenging task and less investigated in a learning-based control society compared to quadrupedal robot locomotion.
Unlike quadrupeds, the humanoids have more complicated structures making it hard to use a straightforward model-free reinforcement learning approach.
In this project, we address this issue by utilizing our established framework, IFM, to enhance the robustness of humanoid locomotion, specifically solving instability in the reinforcement learning stage.
Education
Georgia Institute of Technology
M.S in Electrical and Computer Engineering
Aug. 2022 - Present
Seoul National University
B.S. in Mechanical Engineering B.S. in Computer Science and Engineering
Mar. 2016 - Aug. 2022
Work and Teaching Experiences
CS 8803 Deep Reinforcment Learning for Intel. Control
Teaching Asistant Georgia Institute of Technology
Spring 2024
CS 3451 Computer Graphics
Teaching Asistant Georgia Institute of Technology
Spring 2023
Saige Research
Research Intern
Apr. 2021 - Dec. 2021
Samsung Electronics
Student Intern
Jan. 2021 - Feb. 2021
This template was borrowed from Jon Barron
Last update: Mar.2024