Research Projects
AI-based Multifinger Grasping
Physical AI — learning-based grasping and in-hand manipulation with proprioceptive hardware.
Overview
Bridging hardware and intelligence, we develop learning-based grasping systems that exploit the physical properties of our actuators and sensors — Physical AI in the truest sense.
RL-DOB-based grasping combines reinforcement learning with disturbance-observer control for robust in-hand manipulation. Vision-free blind grasping uses only uniaxial fingertip force sensing to grasp unknown objects without cameras. An ultra-low-impedance gripper enables high-bandwidth, transparent physical interaction.
Demonstrations
Ultra-Low-Impedance Robotic Gripper
Vision-Free Multifingered Blind Grasping
Robust In-Hand Manipulation Policy based on RL-DOB
Ultra-Low-Impedance Robotic Finger — Demo 1
Representative Publications
Learning Blind Grasping with Uniaxial Fingertip Force Sensors
E. Lee, J. Choi, T. Kim, C. Nam*, S. Jeong*
IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM) · 2026
Patents
- Robot grasp control based on uniaxial force sensingFiledKR 10-2026-0014600
Funding
- NRF National Agenda Basic Research — hardware & control core technologies for physical-AI robot hand/gripper platforms (2025.09 – )
- KEIT Robot Industry Core Technology Program — multimodal flexible tactile sensing system for universal grippers (joint)