From force-aware datato learning-baseddexterous bimanual control.

I study how to collect high-quality, physical interaction-aware data and use it to help robots learn human movements more effectively.

I have built a force-feedback leader-arm teleoperation system
designed specifically to collect high-quality demonstration data,
alongside a separate VR-tracker teleoperation system.

I evaluate multiple policies and refine their internal architectures
to improve learning performance, while extending vision–proprioception
policy architectures with force as an additional modality for
learning-based control of dexterous dual-arm robots with hands.

I also developed a ROS 2 visual-servoing package
during an internship with the Robot Control Team at Doosan Robotics.