Loading...

학술행사

Korean AI Association

학술행사

  >   학술행사   >   국내학술대회

국내학술대회

튜토리얼
 
김대겸 교수(고려대학교)
 
Title: 물리인공지능 데이터 구성 및 학습 방법론
 
Abs:
This tutorial explores the concept of Physical AI within the field of robotics. Specifically, it focuses on data collection strategies used in sim-to-real transfer and imitation learning frameworks, discussing their respective strengths and limitations. We will then examine several methods to address these challenges, highlighting both algorithmic and experimental approaches. Finally, I will present some of my ongoing research involving humanoid, wearable, and mobile robots to demonstrate how these ideas are applied in practice.
 

Bio: 

Daekyum Kim received his B.S. degree in Mechanical Engineering from the University of California, Los Angeles, (Los Angeles, CA, USA), in 2015. He earned his Ph.D. degree in Computer Science at KAIST (Daejeon, Republic of Korea), in 2021. He was a Postdoctoral Research Fellow at the John A. Paulson School of Engineering and Applied Sciences, Harvard University (Cambridge, MA, USA), co-affiliated with Wyss Institute. Since September 2023, he has been an Assistant Professor with the School of Smart mobility and the School of Mechanical Engineering, Korea University (Seoul, Republic of Korea). His research interests are in the areas of machine learning, computer vision, robotics, and digital healthcare.

 
김영민 교수 (서울대학교)
 
Title: humanoid motion control and physical reasoning
 
Abs:
Human motion is usually generated by animation engines or captured by motion capture systems. The amount of high-quality motion data is limited, and it is extremely challenging to acquire accurate data for interactive motion. This tutorial explains how we can learn and exploit the latent space of plausible human motions from an existing dataset. We can not only generate diverse human motion, but also combine the architecture with physics simulators to mimic the interactive physics of the motion. Interactive motions can be tailored to the current environment, exploiting the controllability of diffusion models. The tutorial also shares recent advances in exploring latent space representations to enhance the stability of physical reasoning while harmonizing different body parts.
 
Bio: 
Young Min Kim is an Associate Professor in the Department of Electrical and Computer Engineering at Seoul National University, Seoul, Korea, where she is leading a 3D vision lab. She received a B.S. from Seoul National University in 2006 and an M.S. and Ph.D. in electrical engineering from Stanford University in 2008 and 2013, respectively. Before joining SNU, she was a Senior Research Scientist at the Korea Institute of Science and Technology (KIST). Her research interest lies in 3D vision, where she combines computer vision, graphics, and robotics algorithms to solve practical problems. She serves as an area chair for CVPR, ICCV, and ACCV, as a program committee member for Pacific Graphics and AAAI, and as a technical papers committee member for SIGGRAPH and SIGGRAPH Asia. She is also a program chair for 3DV 2026.
 
 
오윤선 교수(한양대학교)
 
Title: LLM/VLM 기반 로봇 계획
 
Abs:
 작업을 수행하기 위해서는 효율적인 계획(Planning) 능력이 필수적이다. 특히 최근에는 대규모 언어모델(LLM)과 비전-언어모델(VLM)의 발전으로 인해, 로봇이 사람의 언어를 이해하고 보다 장기적이며 복합적인 계획을 세우는 것이 가능해지고 있다. 본 튜토리얼에서는 이러한 흐름 속에서 등장한 LLM/VLM 기반의 long-horizon planning 기술을 체계적으로 소개하고, 나아가 로봇이 직접 행동을 생성할 수 있는 다양한 Vision-Language-Action (VLA) 모델을 살펴본다. 이를 통해 참가자들은 최신 로봇 지능 연구의 패러다임 변화를 이해하고, 언어와 시각 정보, 행동을 통합적으로 다루는 차세대 로봇 시스템의 방향성을 모색할 수 있을 것이다.
 
Bio: 

Yoonseon Oh received the Ph.D. degree in the Department of Electrical and Computer Engineering from Seoul National University, Seoul, Korea in 2018. She received the B.S. degree in the Department of Electrical and Electronics Engineering from Seoul National University in 2011.

She is currently an assistant professor, in the department of Electronic Engineering, Hanyang University.  From 2019 to 2021, she was a  Senior Researcher and Researcher at Center for Robotics Research, in Korea Institute of Science and Technology. From 2018 to 2019,  She was a Postdoctoral Researcher in the Department of Computer Science at Brown University, advised by Stefanie Tellex.  She previously received the Ph.D. degree, advised by Songhwai Oh in Robot Learning Laboratory .   Her research interests include robotic intelligence, task learning, and semantic mapping. 

 
 
최준원 교수(서울대학교)
 
Title: 자율주행을 위한 파운데이션 모델 및 VLM 기술
 
Abs:
본 튜토리얼에서는 자율주행 기술이 직면한 주요 도전 과제와 이를 해결하기 위한 핵심 기반 기술들을 종합적으로 살펴본다. 최근 주목받고 있는 End-to-End 자율주행 패러다임의 구조적 특징과 구현 원리를 소개하고, 이를 가능하게 하는 파운데이션 모델의 역할을 논의한다. 특히 리워드 모델을 통한 안전성 강화 전략, 데이터 효율적 학습을 위한 사후학습 기법, 폐회로 시뮬레이션을 위한 월드모델 기반 접근법을 다룬다. 또한 주변 환경의 복잡한 상황을 이해하고 안전한 경로 계획을 수행하기 위해 VLM (Vision-Language Model)의 추론 능력을 활용하는 방법에 대해서도 살펴본다. 마지막으로, 이러한 기술들이 향후 어떤 방향으로 발전하고 통합될지 전망한다.  
 
Bio: 
Jun Won Choi received his B.S. and M.S. degrees from Seoul National University, Seoul, Korea, and his Ph.D. degree from the University of Illinois at Urbana-Champaign, Champaign, IL, USA. From 2010 to 2013, he worked for Qualcomm in San Diego, CA, USA. From 2013 to 2024, he was a Faculty Member in the Department of Electrical Engineering at Hanyang University, Seoul, Korea. Since 2024, he has been a faculty member in the Department of Electrical and Computer Engineering and the Interdisciplinary Program in Artificial Intelligence at Seoul National University, Seoul, Korea. His research interests include signal processing, machine learning, robot perception, autonomous driving, and intelligent vehicles.