Learning-based deformable-object manipulation
Simulation, teleoperation, and policy learning in NVIDIA Isaac Lab.
This research develops a robotic manipulation framework in NVIDIA Isaac Lab for learning grasping and manipulation of irregular, deformable poultry objects from teleoperated demonstrations.
The workflow connects UR10e simulation, multi-view visual observations, robot state, demonstration collection, and diffusion-policy learning with closed-loop evaluation.
Related research interests include imitation learning, deformable-object manipulation, and sim-to-real robotics.