A policy bridge that detects states created by reactive safety interventions and returns the robot to task-relevant behavior. I designed the recovery bridge and evaluated the proposed method in 160 RB10 trials. It completed 146 trials (91.3%); the original LPB reached 64.4% under the same intervention setting.
Hyeonjun Cho
I work on robot learning for physical manipulation under distribution shift. My projects cover imitation learning, vision-language manipulation, and recovery after safety interventions, with most evaluation carried out on a real RB10 manipulator.
Most recently I designed a recovery-aware policy bridge for states created by reactive safety interventions. The proposed method completed 146 of 160 real-robot trials (91.3%), compared with 64.4% reported for the original LPB under the same intervention setting. A manuscript is in preparation.
Research
Force-Feedback Teleoperation for Real-Robot Imitation Learning
RAGTAL / DNA-HERO Team Project
An RB10 imitation-learning system trained from force-feedback teleoperation demonstrations and evaluated through repeated real-robot rollouts. As Learning Lead, I built the demonstration-to-rollout workflow, curated the training data, and conducted 50 real-robot rollouts for each evaluated model.
A tabletop manipulation planner that combines visual grounding, execution memory, and gripper-aware grasp selection for real-robot execution. I connected visual grounding and execution memory to the planner, then evaluated the system on a fixed set of 15 scenarios. Success increased from 9/15 to 12/15.
Zero-Shot Language-to-Motion Planning with Geometric Validation
Reproduction and extension of LMTG
A language-to-motion prototype that converts instructions into waypoint trajectories and checks them for collisions before execution. I implemented the task representation, waypoint parser, and collision checks, then connected the planner to simulation and initial real-robot trials.
Awards
- Grand Prize, Creative Innovation DNA-HERO Industry-Academia Team Project, Sungkyunkwan University stage (2025).
- Silver Award, Consortium Creative Capstone Design Competition, National Preliminary Round (2025).
Profile
- Education
- B.S. in Mechanical Engineering, expected February 2027; Double Major in Physics, expected August 2027, Sungkyunkwan University.
- Methods
- Imitation learning, vision-language models, motion planning.
- Tools
- Python, PyTorch, ROS 2, real-robot systems (RB10).