Seunghwan Jang

M.S. Student, School of Electrical Engineering, KAIST jsh991124@kaist.ac.kr

M.S. student in the School of Electrical Engineering at KAIST (Autonomous Control for Stochastic Systems Lab; advisor: Prof. SooJean Han).
Research interests include safe generative motion planning (flow matching / diffusion + control barrier functions) and efficient 3D Gaussian Splatting (3DGS)-based perception/situational awareness systems.

I am always open to discussing research and enjoy exchanging new ideas. Please feel free to reach out to me anytime!

Education

Korea Advanced Institute of Science and Technology (KAIST)

M.S. Student, Department of Electrical & Electronic Engineering
Autonomous Control for Stochastic Systems Lab (Daejeon, Korea)
Advisor: Prof. SooJean Han
Sep 2024 – Present

Kyung Hee University

B.S. in Computer Science & Engineering
College of Software Convergence (Yongin, Korea)
Advisor: Prof. Hui Yong Kim
GPA: 4.02 / 4.3 (Summa Cum Laude; Ranked 1st/97)
Mar 2019 – Aug 2024

Experience

M.S. Student Researcher

Autonomous Control for Stochastic Systems Lab, KAIST (Daejeon, Korea)

• SafeFlowMatcher: flow-matching trajectory generation + Control Barrier Function (CBF-QP) safety filtering for constraint-satisfying planning.
• 3D Gaussian Splatting (3DGS) pipeline for situational awareness / visual sensor networks (AFOSR), focusing on deployment constraints (latency, bandwidth).

Sep 2024 – Present

Research Student

Visual Media Lab, Kyung Hee University (Yongin, Korea)

• Built and evaluated MPEG FCM/FCVCM pipelines combining conventional codecs (VVC, HEVC) with neural compression; emphasized rate–accuracy trade-offs and reproducibility.
• 3 first-author domestic publications; 10 MPEG standardization contributions; 1 Korean patent.

Aug 2022 – Aug 2024

Research Intern

Electronics and Telecommunications Research Institute (ETRI) (Daejeon, Korea)

Supported MPEG standardization by running validation experiments and analyzing performance data for FCM/FCVCM.

Jul 2023 – Aug 2023

Research Project

AFOSR (United States)

Fast, accurate, and efficient situational awareness using visual sensor networks (3D Gaussian Splatting–based system).

Jun 2025 – Present

Research

Focus Areas
  • Safe generative motion planning (flow matching / diffusion-style models + Control Barrier Functions)
  • Deployment-aware perception and situational awareness with 3D Gaussian Splatting (3DGS) and visual sensor networks

Keywords: safe planning, flow matching, diffusion models, control barrier functions, CBF-QP, 3D Gaussian Splatting, 3DGS, visual sensor networks

Publications
Stratified Hazard Sampling: Minimal-Variance Event Scheduling for CTMC/DTMC Discrete Diffusion and Flow Models
Seunghwan Jang, SooJean Han
arXiv | 2026-01 | arXiv preprint
arXiv GitHub
SafeFlowMatcher: Safe and Fast Planning using Flow Matching with Control Barrier Functions
Seunghwan Jang*†, Jeongyong Yang*, SooJean Han
arXiv | 2025-10 | Accepted at ICLR 2026; arXiv preprint
Equal contribution (*); corresponding author (†) on CV.
arXiv OpenReview GitHub
Comparison of Compression Codec Algorithms to Image Datasets in 3D Gaussian Splatting Models
Seunghwan Jang, SooJean Han
Workshop on Image Processing and Image Understanding (IPIU) | 2025-02 | Poster (in Korean)