cv

Graduate Research Assistant

Hanyang University

Robot learning research under Prof. Youngmoon Lee.

• Self-Supervised Critical Phase Detection (CPD): step-level premature-commitment detector and ≈ 1K-parameter residual adapter for frozen π₀ / π₀.₅ VLA policies on LIBERO; cross-modal execution gap between action-expert latent and proprioception (preprint, under review).
• Poisoning attacks on multi-agent RL: reward-poisoning attacker agent on cooperative PPO/SAC crawlers in a Unity 50×50 m benchmark; 18.7%/20.9% reward drop in multi-agent PPO/SAC, up to 98.1% in single-agent SAC (Humanoids 2025 LBR).
• Safety-RL: bridging the sim-to-real gap by allowing safe optimization and exploration during robot learning.
• Vision (HumanPose): hybrid PoseNet/SegNet approach for crowded scenes with overlap and occlusion.

Research Intern

KITECH — Process Platform Research Division

• Researched AI methods for detecting defective products on high-speed conveyor belts.

Research Intern

Hanyang University — CAI Lab

• IKEA Furniture Assembly Collaborative Robot: RL simulation in which two manipulators identify randomly placed furniture parts and assemble them collaboratively.

MS

Hanyang University

Applied Artificial Intelligence

• Advisor: Prof. Youngmoon Lee
• Research on safety-RL, poisoning attacks on multi-agent RL, and vision-language-action policy failure detection.

BS

Hanyang University

Robotics

• Advisors: Prof. Youngmoon Lee, Prof. Taejoon Park, Prof. Changsoo Han

Self-Supervised Critical Phase Detection for VLA Refinement

Preprint (under review)

Self-supervised step-level critical-phase detector and ≈ 1K-parameter residual adapter that refines frozen π₀ / π₀.₅ VLA policies inside detected phases.

Poisoning Attacks on Multi-Agent Reinforcement Learning Systems

IEEE-RAS International Conference on Humanoid Robots (Humanoids), Late-Breaking Report

HAPtics: Human Action Prediction in Real-time via Pose Kinematics

International Conference on Pattern Recognition (ICPR), Kolkata, India

Causes and Fixes of Unexpected Drone Shutoffs

ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED)

Snapbot: Enabling Dynamic Human Robot Interactions for Real-Time Computational Photography

ACM/IEEE International Conference on Human-Robot Interaction (HRI), Late-Breaking Report

Leveraging Keypoints as Dynamic Centroids for Unified Representation of Human Pose and Instance Segmentation

Preprint (withdrawn from CVPR 2024)

SSK-DNN: Semantic and Sentiment Knowledge for Incremental Text Sentiment Classification

IEEE ICDM Workshop on Incremental Learning (IncrLearn), Shanghai, China

Information & Communication Planning and Evaluation President Award (2nd place) — SW Talent Festival

AnimeGAN Filter Photographer: a manipulator that frames and captures stylized portraits under safety-RL constraints.

Capstone Design Award (3rd place) — Hanyang University

Guide Robot Dog: a full-stack quadruped robot motivated by guide dogs, intended to assist visually impaired pedestrians.