Scaling robotic manipulation with physics simulators & web-scale data — toward human-level manipulation intelligence
I am a robotics AI researcher focused on scaling robotic manipulation with large-scale physics simulators and web-scale data. My goal is to realize human-level object manipulation. Recently, I have been exploring Vision-Language-Action model (VLA) systems, Reinforcement Fine-Tuning (RFT), Imitation Learning, 3D point cloud perception, and tactile sensing for robust manipulation policies.
- Ph.D. Student, Robotics Intelligence Program, GIST — Sep 2022 – Present
- M.S., Robotics Intelligence Program, GIST — Sep 2020 – Aug 2022
- Thesis: Learning-based 3D Point Cloud Analysis and Applications for Robotic Manipulation in the Real World
- B.S., Mechanical Engineering, Dankook University — Mar 2015 – Aug 2019
- GraspSAM: When Segment Anything Model Meets Grasp Detection, ICRA 2025
- Learning to Place Unseen Objects Stably Using a Large-Scale Simulation, IEEE RA-L — 29th Samsung Humantech Paper Award, Gold Prize
- GraspClutter6D: A Large-Scale Dataset for Robotic Grasping and Perception in Clutter, IEEE RA-L
- PolyFit: A Peg-in-Hole Assembly Framework for Unseen Polygon Shapes via Sim-to-Real Adaptation, IROS 2024
- Unseen Object Amodal Instance Segmentation via Hierarchical Occlusion Modeling, ICRA 2022
For a complete list and details, see my Google Scholar.
- Apparatus and Method Identifying the Size of the Target Object, US No. 12,380,586 (2025)
- Method for Generating a Grasp Map Based on Object Segmentation and Learning Method, KR No. 10-2567558 (2024)
- Method and System for Estimating a Stable Plane to Place an Object in a Stable Pose, KR (2024)
- Plus others
- Outstanding Paper Award, ICROS-KROS 2025
- Gold Prize (Top 0.7%), 29th Samsung Humantech Paper Award — Samsung Electronics ($10,000)
- Outstanding Paper Award, ICROS-KROS 2022
- Best Robot Vision Paper Award, AFCV @ KROC 2022
- Email: sangjun7@gm.gist.ac.kr
- Location: South Korea