Seungyeon Yoo

I am a Ph.D. candidate at Seoul National University, advised by Prof. H. Jin Kim. My research interests include robot learning, vision-based navigation, and visual representation learning.

Email  /  Resume  /  LinkedIn

Research

My research focuses on building RGB-only autonomous robot systems. I have been exploring photorealistic real-to-sim data generation for robust sim-to-real transfer of RGB-only policies, monocular target chasing via cross-modal representation learning, and single-view manipulation with 3D-aware representation learning. I apply these approaches on drones, manipulators, and ground robots. I am currently extending real-to-sim simulation to humanoid loco-manipulation, in collaboration with NVIDIA Research. I believe that RGB-focused robot systems can serve as a key pathway toward realizing human-level intelligence in robotics. Key publications are highlighted below.

ReaDy-Go generates photorealistic navigation datasets for dynamic environments by combining a reconstructed static GS scene with dynamic human GS obstacles, and trains policies robust to both the sim-to-real gap and moving obstacles.

BooST learns a unified skill representation that bridges semantic intent (what) and motion dynamics (how), then distills it into a lightweight policy — enabling few-shot adaptation across new scenes, tasks, and even cross-embodiments, with robustness to dynamic visual distractors.

We propose a novel RL framework that extracts 3D-aware representations from single-view RGB input, without requiring camera pose or synchronized multi-view images during the downstream RL.

Drift in visual pose estimation is eliminated through plane-based joint optimization and the registration module.

Cross-modal representation enables RGB-only target chasing instead of multiple sensor inputs.

This page was built with Dr. Jon Barron's template.