I am a mathematics-trained robotics researcher interested in real-world robot learning, VR teleoperation, imitation learning, and high-frequency control. My work focuses on turning research ideas into reliable real-robot systems.
I developed the real-time integration layer that converts third-party teleop_xr IK targets into filtered, rate-limited, and safety-monitored DaMiao MIT CAN commands for a physical 14-DOF OpenArm platform.
Code and documentation · Primary object-manipulation demo · Previous VR demo
VR position/orientation mapping, MuJoCo motion validation, and real-time UR5e control using a 25 Hz VR input and a 125 Hz RTDE control loop.
Code and documentation · Full demo
Real-robot reproduction of an ACT-based imitation-learning pipeline for delicate grape grasping. This is presented as a video-only reproduction demo because the original training code and dataset are no longer available.
Python · PyTorch · MuJoCo · UR RTDE · CAN / CAN FD · VR teleoperation · Imitation learning
Beijing, China · yaoruliu@bjfu.edu.cn


