Research

Ongoing graduate research in dynamics modeling and whole-body control, advised by Prof. Shinsuk Park.

Hold Before You Step: IMU-Resolved Disturbance Estimation for In-Place Push Recovery on Force-Sensorless Humanoids

2026 · Humanoid Robotics · Disturbance Estimation · Preprint (arXiv)

Measures how fast and accurate a sensorless humanoid's push estimate needs to be to safely hold in place instead of stepping — then uses that to cut base displacement by 96% on real pushes.

Role: First Author

Dynamics Simulation Parameter Tuning for Sim-to-Real Trajectory Similarity of a Reinforcement Learning Policy for LiDAR-based Mapless Navigation

2026 · Mobile Robotics · Sim-to-Real RL · Accepted, Journal of ICROS

Tunes a simulator's wheel-friction and suspension parameters to real robot response data, cutting the sim-to-real trajectory gap of a LiDAR mapless-navigation RL policy by double digits — cross-validated with two independent trajectory-estimation methods.

Role: Co-author

Feasible Is Not Faithful: Intent-Preserving Constraint Realization for Contact Tasks on Unknown Surfaces

Sep.2025 -- Present · Robot Learning · Contact-Rich Manipulation · Preprint (arXiv) · Submitted to IEEE RA-L

Master's Research Encouragement Grant, National Research Foundation of Korea (2025)

A control framework where the task specifies what to preserve under unknown contact — on tilted boards, it kept a drawn figure within 0.5 mm versus 10–18 mm for the standard nearest-point rule.

Role: First Author

Transition-Stable Hierarchical Quadruped Control via Learned Latent Motion Representations

2025 -- 2026 · Legged Robotics · Reinforcement Learning · Working paper

Adapts CALM's latent motion representations to quadruped control for transition-stable locomotion and object interaction.

Role: Co-author

Real-Time Architecture for a Generative-AI-Based Mobile Manipulator

Mar.2025 -- Aug.2026 · Systems Integration · Mobile Manipulation · Ongoing

A ROS2 architecture connecting simulation, an LLM, a UI, and a physical mobile manipulator, with edge-based data collection and automatic cloud sync that preserves the robot's real-time control cycle.

Role: System Architect

Balance Demand Descriptor: A Physics-Informed Representation of Whole-Body Dynamics for Humanoid Standing Manipulation

Mar.2025 -- Present · Humanoid Robotics · Whole-Body Control · Manuscript in preparation

A physics-informed Balance Demand Descriptor, derived from centroidal dynamics, that lets a humanoid's control policy explicitly distinguish disturbances it can absorb in place from those requiring a corrective step.

Role: First Author

Earlier Projects

Undergraduate coursework and competition projects in mechanism design and robot perception.

Quadruped Robot Obstacle Avoidance via ORB-SLAM and UV-Disparity

Mar.2024 -- Jan.2025 · Computer Vision · Autonomous Navigation

President's Award (1st Place), Creative Challenger Program (CCP), Korea University — Jan.2025

Monocular ORB-SLAM depth estimation and UV-disparity segmentation for real-time obstacle avoidance on a quadruped, without a dedicated depth sensor.

Role: Team Leader, RoboStride

Solar Tracking Parasol

Sep.2023 -- Nov.2023 · Mechanism Design and Control

Bronze Award, Creative Comprehensive Design Competition — Sep.2023; Design Award, Capstone Design Fair, Korea University — Nov.2023

A parasol that senses sun position and actuates its canopy to track the sun and maximize shade throughout the day.

Role: Team Leader

Ladder-Crossing Robot

Mar.2023 -- Sep.2023 · Mechanism Design and Control

A tendon-driven, 3D-printed robot that climbs a horizontal ladder by gripping rungs and shifting its center of mass.

Role: Team Leader