Seonwoo Kim

Seonwoo Kim

MS in Mechanical Engineering, Korea University

whimoon1826@naver.com ssenukim.github.io LinkedIn
Download CV

About

I am a robotics engineer specializing in robotic manipulation. I received both my B.S. and M.S. degrees in Mechanical Engineering from Korea University. During my Master's studies, I built a strong foundation in classical control engineering and state estimation theory for robotic manipulators. Currently, my research focuses on humanoid bi-manual manipulation using data-driven approaches. My ultimate goal is to address complex manipulation tasks by integrating classical model-based control and planning methods with modern learning-based techniques.

Current Interests

My research interests lie in contact-rich robot manipulation and lightweight robot learning models. Currently, I am exploring the following key questions:

  • How can neural networks effectively comprehend and represent contact dynamics?
  • How can we address the data scarcity problem in robot learning?
  • How can we integrate dynamics-based control with data-driven algorithms to accelerate inference speed without compromising performance?

I believe that dexterous manipulation is a prerequisite for robots, including humanoids, to seamlessly coexist and collaborate with humans in everyday environments. In this context, my ultimate goal is to develop safe, human-level robot intelligence that enables humanoids to work alongside people.

Education

Korea University

MS in Mechanical Engineering

Sept 2023 – Aug 2025
  • Advisor: Prof. Daehie Hong
  • GPA: 4.5/4.5
  • Thesis: Adaptive Contact Force Estimation for Robotic Manipulators Using an Interacting Multiple Model-Based Disturbance Kalman Filter and Bayesian Neural Network
  • Coursework: Linear system, Vehicle dynamics, Robust control, Intelligent control, Digital control, Optimal control, Kinematics Analysis

Korea University

BS in Mechanical Engineering

Mar 2017 – Aug 2023
  • GPA: 3.55/4.5
  • Coursework: Robotics, Mechatronics, Kinematics

Teaching Experience

Teaching Assistant, MECH226 Dynamics

Korea University, Prof. Daehie Hong

Mar 2024 – Dec 2024

Teaching Assistant, MOBI208 Dynamics

Korea University, Prof. Seongho Yun

Mar 2025 – Jun 2025

Publications

Journal Articles

Task Space End-Effector Contact Force Estimation for Robotic Manipulators Using a Bayesian Augmented Interacting Multiple Model-Based Disturbance Kalman Filter

Seonwoo Kim, Myeongin Jin, Jihun Kim, Chanwoo Kim, Daehie Hong

IEEE Transactions on Instrumentation and Measurement (IEEE TIM), 2026 (Accepted)

SCIE, First author

Acceleration Measurement-Free Dissipative Disturbance Observer for Robotic Manipulators

Seonwoo Kim, Chanwoo Kim, Yeonho Ko, Daehie Hong

IEEE Robotics and Automation Letters (IEEE RA-L), 2025

SCIE, First author

DOI: 10.1109/LRA.2025.3564205

Conference Proceedings

Soil Interaction Force Estimation via Identified Excavator Dynamics with Particle Filter

Seonwoo Kim, Jihun Kim, Chanwoo Kim, Daehie Hong

International Conference on Precision Engineering and Sustainable Manufacturing (PRESM 2024)

Estimating the Interaction Force Between Excavator Bucket and Ground Using a Simplified Dynamical Model and Hydraulic Sensors

Seonwoo Kim, Jangho Bae, Chanwoo Kim, Jihun Kim, Daehie Hong

Proceedings of the KSPE 2024 Spring Conference

Optimal Excavator Bucket Trajectory Generation with Soil Mechanics Model Design

Seonwoo Kim, Jangho Bae, Jaemyung Huh, Chanyoung Moon, Cheolhwan Im, Jinwoo Park, Daehie Hong

Proceedings of the KSPE 2023 Autumn Conference

Research Experience

Korea Institute of Science and Technology (KIST)

Intern Researcher

Jul 2025 – Dec 2025
  • Autonomous bi-manual manipulation for humanoid robots: Developed humanoid bi-manual manipulation policies by training robot foundation model based AI model (Nvidia Gr00t) with combined real-world and synthetic datasets, enabling robust dual-arm coordination

    Demo Videos:

    Robot foundation model deployment on humanoid platform

    Synthetic data generation for training bi-manual manipulation policies

  • XR-based humanoid teleoperation system: Developed an integrated teleoperation system combining Meta Quest hand tracking, camera/sensor modules, and Jetson Orin-based control architecture to enable real-time humanoid robot manipulation through inverse kinematics

    Demo Videos:

    Boxing demonstration with bilateral arm coordination

    Contact-rich door opening task with precise manipulation

    Dexterous bi-manual water pouring with fine motion control

    Block stacking task demonstrating spatial coordination

  • Reinforcement learning for dexterous manipulation: Trained humanoid dexterous manipulation policies using reinforcement learning in NVIDIA Isaac Sim/Lab simulation environment for complex object manipulation tasks

    Demo Video:

    RL training process in NVIDIA Isaac Sim for dexterous manipulation

Korea University

Student Researcher

Sep 2023 – Aug 2025

AI-based autonomous excavator control: Imitation learning based trajectory generation policy design using LSTM networks to learn from expert operators, enabling autonomous execution of diverse excavation tasks (radial digging, trenching, slope work, and stockpile excavation)

Demo Videos:

Radial Excavation

Stockpile Excavation

6-DOF motor grader blade control system: Designed control system for motor grader blade (hybrid 3-DOF serial + 3-DOF parallel manipulator), implementing forward/inverse kinematics solvers and pose control algorithms for autonomous blade positioning

External force estimation for construction equipment: Developed disturbance observer and Kalman filter-based contact force estimator to estimate soil-bucket interaction forces without additional force/torque sensors (published in IEEE RA-L and TIM)

Awards

MATLAB Korea AI Competition for University Students

2023

Bronze Award (Team Leader)

  • Led a team to develop an autonomous excavator trajectory generation system using Deep Learning
  • Implemented a GRU-based network in MATLAB to clone expert excavation skills from human demonstration data
  • Integrated Intel RealSense D455 for real-time terrain perception and achieved smooth digging motion control on a 1/14 scale excavator

Technologies & Skills

Programming Languages

C, C++, Python, Matlab/Simulink, Labview, Julia

Developer Tools

PyTorch, Isaac Sim/Lab, ROS2, Git, Docker

CAD

SolidWorks