profile photo

Seungyeon Kim

Hello! I am postdoctoral researcher at Robotics Laboratory in Seoul National University, advised by Frank C. Park.

Prior to joining postdoctoral program, I completed my Ph. D., M.S., and B.S. with Mechanical Engineering (minor: Economics) from Seoul National University.

Email: ksy at robotics.snu.ac.kr

CV   /  Google Scholar   /  Github   /  Youtube


Research (Research Statement)

My research focuses on developing practical solutions for intelligent robots that are adaptive and generalizable to unknown arbitrary environments. I am particularly interested in leveraging inductive biases to enable robots to perform effectively in real-world scenarios with limited data while also adapting to various downstream manipulation tasks. My research aims to address the challenges of designing effective inductive biases for intelligent robots and contribute to the development of more efficient and capable robotic systems.

  • Introducing inductive bias into the recognition system enables efficient 3D reconstruction of objects from partial and incomplete visual observations, allowing robots to manipulate them effectively [DSQNet, T2SQNet, Search-for-Grasp].
  • Adopting equivariant models reduces data requirements while enhancing generalizability across diverse robot manipulation tasks, ranging from pushing dynamics learning to skill learning [SQPDNet, EMMP].
  • Identifying low-dimensional representations of robot trajectories reduces the complexity of high-dimensional data, enabling rapid adaptation to environmental changes [EMMP, DIVO].

Publications ( * denotes equal contribution.)
ms
Diverse Policy Learning via Random Obstacle Deployment for Zero-Shot Adaptation
Seokjin Choi*, Yonghyeon Lee*, Seungyeon Kim, Che-Sang Park, Himchan Hwang, Frank C. Park
IEEE Robotics and Automation Letters (RA-L) 2025
Project Page   •   Paper   •   Bibtex

ms
T2SQNet: A Recognition Model for Manipulating Partially Observed Transparent Tableware Objects
Young Hun Kim*, Seungyeon Kim*, Yonghyeon Lee, Frank C. Park
Conference on Robot Learning (CoRL) 2024
Project Page   •   Paper   •   Code   •   Bibtex

ms
Leveraging 3D Reconstruction for Mechanical Search on Cluttered Shelves
Seungyeon Kim*, Young Hun Kim*, Yonghyeon Lee, Frank C. Park
Conference on Robot Learning (CoRL) 2023
Project Page   •   Paper   •   Code   •   Bibtex

emmp
Equivariant Motion Manifold Primitives
Byeongho Lee*, Yonghyeon Lee*, Seungyeon Kim, MinJun Son, Frank C. Park
Conference on Robot Learning (CoRL) 2023
Project Page   •   Paper   •   Code   •   Bibtex

se2
SE(2)-Equivariant Pushing Dynamics Models for Tabletop Object Manipulations
Seungyeon Kim, Byeongdo Lim, Yonghyeon Lee, Frank C. Park
Conference on Robot Learning (CoRL) 2022
Oral Presentation   •   Project Page   •   Paper   •   Code   •   Supplementary   •   Bibtex

dsqnet
DSQNet: A Deformable Model-Based Supervised Learning Algorithm for Grasping Unknown Occluded Objects
Seungyeon Kim*, Taegyun Ahn*, Yonghyeon Lee, Jihwan Kim, Michael Y. Wang, Frank C. Park
IEEE Transactions on Automation Science and Engineering (T-ASE) 2022
Project Page   •   Paper   •   Code   •   Bibtex

smf
A Statistical Manifold Framework for Point Cloud Data
Yonghyeon Lee*, Seungyeon Kim*, Jinwon Choi, Frank C. Park
International Conference on Machine Learning (ICML) 2022
Paper   •   Code   •   Bibtex

jnp
On the Encoding Capacity of Human Motor Adaptation
Seungyeon Kim, Jaewoon Kwon, Jin-Min Kim, Frank C. Park, Sang-Hoon Yeo
Journal of Neurophysiology (JNP) 2021
Paper   •   Bibtex

Projects
sr
Object Grasping and Manipulation Skills for Stable Housekeeping Service
Project Leader
The goal of the project is to develop skills to enable various household tasks, specifically, to develop a skill that can grasp various tableware and objects on the table.

lane
Deep Learning-based Lane Detection Algorithm from LiDAR data
Project Leader
The goal of the proejct to develop a neural network architecture that recognizes 3D lane information from LiDAR data (3D point cloud + point-wise intensity).

drl
Deep Reinforcement Learning Algorithm for Industrial Robot
Project Leader
The goal of the proejct is to develop safe and efficient reinforcement learning algorithm for high-gain position controller-based industrial robots.

se2
Artificial Intelligence-based Automated Painting Robot System
Project Member (Role: Visualization)
The goal of the project is to develop an artificial intelligence-based smart painting robot automation system for automobile factories.


Education

  • Feb 2024. Ph.D. degree in Mechanical Engineering, Seoul National University
    • Advisor: Prof. Frank Chongwoo Park
    • Thesis: Learning for Vision-Based Object Manipulation: A Shape Recognition-Based Approach
         PDF   •   Slides
    • Outstanding Doctoral Dissertation Award
  • Feb 2019. M.S. degree in Mechanical Engineering, Seoul National University
    • Advisor: Prof. Frank Chongwoo Park / work closely with Prof. Sang-Hoon Yeo
    • Thesis: On the Encoding Capacity of Human Motor Adaptation
  • Feb 2017. B.S. degree in Mechanical Engineering and Minor in Economics, Seoul National University
    • Summa Cum Laude
  • Feb 2013. Gyeonggibuk Science High School
    • One-year early graduation

    Teaching Experiences

  • Teaching Assistant, Geometric Methods for High-Dimensional Data Analysis (M3239.006800), SNU, 2022F
  • Teaching Assistant, Dynamics (446.204A), SNU, 2018F
  • Teaching Assistant, Introduction to Robotics (M2794.0027), SNU, 2017S
  • Undergraduate Student Instructor, Basic Calculus 1 (033.016), SNU, 2015S
  • Undergraduate Student Instructor, Basic Calculus 2 (033.017), SNU, 2014F

  • Template based on Jon Barron's website.