About

I am an M.S. student in the Interactive Robotics Lab at Dongguk University, advised by Prof. Soo-Chul Lim. My research focuses on bimanual robot learning and control, with a particular emphasis on Vision-Language-Action (VLA) models and their application to real-world manipulation tasks.

Research Interests

  • Bimanual Robot Learning & Control
  • Vision-Language-Action Models
  • Robot Teleoperation
  • Mixture-of-Experts Architecture
  • Multimodal Perception for Robotics

Education

M.S. in Mechanical Engineering
Dongguk University, Seoul — Interactive Robotics Lab (Advisor: Prof. S.-C. Lim)
Sep 2025 – Present
B.S. in Mechanical, Robotics & Energy Engineering
Dongguk University, Seoul
Mar 2022 – Aug 2025

Publications

[P1]
See Selectively, Act Adaptively: Dual-Level Structural Decomposition for Bimanual Robot Manipulation
Y.-J. Choi, Y.-C. Son, and S.-C. Lim
arXiv:2606.13279, 2026 · Project Page Under Review

See Selectively, Act Adaptively decomposes bimanual robot manipulation into view-selective perception and interaction-aware action routing to improve robust, long-horizon task performance.

[J1]
ThermoAct: Thermal-Aware Vision-Language-Action Models for Robotic Perception and Decision-Making
Y.-C. Son, D.-K. Ko, Y.-J. Choi, and S.-C. Lim
IEEE Robotics and Automation Letters (RA-L) · Project Page Accepted

A hierarchical VLM–VLA framework that leverages thermal information to adapt task plans to environmental conditions, enabling safer, temperature-aware robotic manipulation.

[P2]
Task-Aware Modality Selection for Multimodal VLA in Robotic Manipulation
Y.-C. Son, J.-W. Lee, Y.-J. Choi, D.-K. Ko, and S.-C. Lim
arXiv preprint arXiv:2602.15543, 2026 · Project Page Under Review

A lightweight router for a pretrained VLA that dynamically emphasizes task-relevant sensory streams while attenuating irrelevant ones.

Research Projects

Humanoid Dexterous Manipulation with RBY1 2026
RBY1 · Inspire Hand · VLA · Sim-to-Real
  • Trained VLA policies for everyday tasks such as pick-and-place and pouring
  • Leveraged both real-world demonstrations and simulation-augmented data for training
UR5e Bimanual Teleoperation System 2026
UR5e · Meta Quest VR · MuJoCo
  • Real-time teleoperation via Meta Quest VR controllers
  • Inter-robot collision avoidance using MuJoCo geometric distance computation
Fine-tuning VLA Models (pi0.5, GR00T N1.7) 2025 – 2026
PyTorch · pi0.5 · GR00T N1.7
  • Sensor-grounded robot task fine-tuning for manipulation scenarios
Mobile Manipulator Remote Control GUI 2024
Scout Mini · ROS · Python
  • Integrated system GUI for mobile manipulator control (Scout Mini + robotic arm)
Autonomous Hamburger Assembly Robot 2024
OpenCV · Collaborative Robot
  • End-to-end vision-based robotic assembly using collaborative arm

Experience

M.S. Student Sep 2025 – Present
Interactive Robotics Lab, Dongguk University
Bimanual robot learning and control; developing Mixture-of-Experts VLA architecture. Building teleoperation systems for dual UR5e arms with VR-based control interfaces.
Team Leader — SOOMAC (Robot Arm Team) 2023 – 2024
Dongguk University
Led computer vision development for robotic arm manipulation tasks; coordinated team of engineers for competitive robot arm challenges.
  • Minister of Trade, Industry and Energy Award, Creative Engineering Design Competition, 2024
  • Best Award, IoT Home Appliance Consortium Engineering Design, 2024
  • and 4 more
Member — MACARON (Autonomous Driving Team) 2022 – 2024
Dongguk University
Path planning and tracking algorithms for autonomous vehicle systems; participated in multiple national autonomous driving competitions.
  • Grand Prize, Future Mobility Urban Challenge, 2022
  • Innovation Award, SNU Autonomous Driving SW Challenge, 2023
  • and 2 more