Event
Proposes FRDV fast retargeting and dynamics validation framework
1 reports · 1 sources
Daniel Marew et al. proposed the FRDV framework, which rapidly retargets large-scale human motion capture data to different humanoid robots and performs dynamics validation on the required motions. The paper claims a retargeting speed of up to 9,156 frames per second, approximately 136 times that of GMR and ProtoMotions; for motions requiring additional validation, a contact-implicit iLQR controller is used to verify the retargeted references under full rigid-body dynamics, and sim-to-real transfer was demonstrated on the Unitree G1 humanoid robot.
Report timeline
- FRDV: A Fast Motion Retargeting Framework with Dynamics Validation for Humanoid Robots
The paper proposes FRDV (Fast Retargeting with Dynamics Validation), a framework for rapidly retargeting large-scale human motion capture data to different humanoid robots.
arXiv cs.RO ·