Skip to content

Motion Retargeting Example

This entry applies to DR02 Pro. deep-robotics-retarget uses inverse kinematics to convert human motion into robot joint motion, with MuJoCo visualization.

Preparation and workflow

Section titled “Preparation and workflow”
  1. Follow the repository guide to prepare Python, the DR02 Pro model, and the required human models/motion data. SMPL-X, AMASS and BVH inputs have different preparation requirements.
  2. Select the conversion example for your input format and produce robot motion files. Inspect a single motion before batch processing.
  3. View the result in MuJoCo and inspect scaling, joint poses, contacts and collisions. Retargeted motion is reference data, not a command stream ready to send to hardware.
  4. To train a tracking policy, continue to the imitation-learning example. Follow that repository’s current format and forward-kinematics conversion requirements.
Section titled “Related resources”