Junsung Park

Undergraduate student at Seoul National University

I am an undergraduate student at Seoul National University (SNU), majoring in Electrical and Computer Engineering with a minor in Mechanical Engineering.

I am currently a researcher at CONFIG, advised by Kimin Lee and Hyungmok Son. Previously, I conducted research with Xiaolong Wang and Sha Yi at UC San Diego. At SNU, I have also worked with Frank Chongwoo Park and Yong-Lae Park.

GPA: Seoul National University: 3.89/4.0  |  UC San Diego: 4.0/4.0

I am applying to PhD programs for Fall 2027.

I study how robots can leverage human sensory modalities, such as vision and tactile, to learn human-like manipulation.

I aim to enable robots to learn physical dynamics through interaction. By coupling active perception with uncertainty-aware planning and control, I study how robots can use tactile, force feedback, and proprioception to refine their models and adapt their actions in real time. Ultimately, I seek to turn accumulated experience into generalizable physical knowledge for human-like manipulation, with the broader goal of enabling robots to execute complex tasks and assist people.

Publications

Click any image or video to enlarge.
Semi-Supervised Inverse Dynamics Models: Labeling Human Videos with Robot Actions
Junsung Park, Jaehwi Song, Byeongguk Jeon, Suchae Jeong, Sungdong Kim, Minjoon Seo, Hyungmok Son, and Kimin Lee
CVPR 2027, under review
TACTX: Learning Shared Tactile Representations Across Diverse Sensors
Junsung Park*, Sachin Bhadang*, Carmelo Sferrazza, Sha Yi, and Xiaolong Wang
CoRL 2026
paper / project page / dataset
Modality-Augmented Fine-Tuning of Foundation Robot Policies for Cross-Embodiment Manipulation on GR1 and G1
Junsung Park, Hogun Kee, and Songhwai Oh
RSS 2026, Workshop: Tactile Sensing for Robotic Foundation Models
paper
TacO: Benchmarking Tactile Sensors for Object Manipulation
Anya Zorin, Zilin Si, Myungsun Park, Junsung Park, Alexiy Buynitsky, Sachin Bhadang, Taejun Park, Sohee John Yoon, Yong-Lae Park, Oliver Kroemer, Zeynep Temel, Michael T. Tolley, Sha Yi, and Xiaolong Wang
Preprint, 2026
paper

Projects

CEC and tabular GPI tracking a figure-eight past four obstacles on the same noise
Safe Trajectory Tracking under Motion Uncertainty
Junsung Park
Course project, continued after UCSD ECE 276B, 2026
project page
Compared receding-horizon CEC with policy iteration under motion noise, and built a CEC–GPI hybrid with path-aware collision risk that tracks closer while staying safe.
Wearable Rehabilitation Glove
Junsung Park, Injae John, Sungjae Hwang, Yong-Lae Park
Soft Robotics & Bionics Lab (SRBL), Seoul National University, 2024
Developed a wearable rehabilitation glove for stroke patients, integrating a rack-and-pinion actuator, e-gain soft sensor, and IMU-based feedback control to provide adaptive thumb motion assistance.
Unitree G1 Multimodal Manipulation Dataset
Junsung Park
Open Source Project & Dataset, 2025
dataset
Built a large-scale multimodal dataset-generation pipeline integrating Robosuite, RoboCasa, and cuRobo, producing synchronized RGB-D, contact-force, and proprioceptive data for training foundation robot policies.
Retargeting Human Motion to Unitree G1
Junsung Park
Humanoid Imitation Learning Project, 2025
Developed a whole-body imitation learning pipeline that retargets human–object interaction motions from the OMOMO Dataset to the Unitree G1 using SMPL-based preprocessing and gradient optimization.
Active Rainwater Removal System
Junsung Park, Hyuntae Hwang, Seungha Woo, Minjae Lee
Mechatronics Design Competition, Awarded, 2023
Developed a centrifugal umbrella system that actively removes rainwater using a high-torque motor and rotational airflow. Led the system design and integration, achieving fast water removal, low noise, and portable operation.
Bird’s-Eye View Semantic Mapping in AI2-THOR
Junsung Park
Scene Understanding & Mapping Project, 2025
Built a BEV mapping pipeline in AI2-THOR that converts egocentric RGB-D into depth-based semantic top-down maps. Implemented a Trans4Map-style egocentric-to-allocentric model for accurate online BEV scene reconstruction.

Contact

You are welcome to contact me regarding my research! My email is night1115 [at] snu [dot] ac [dot] kr.