Disney Research’s Differentiable Optimal Control (DOC) was designed to help legged robots walk with rich motions. This framework facilitates the transfer of rich motions from either animals or animations onto these robots.
The team used motion capture or animation data to formulate re-targeting objectives whose parameters make them question the differences in proportions and numbers of degrees of freedom between input and robot. By optimizing these parameters over the manifold spanned by optimal state and control trajectories, re-targeting errors are minimized.
No products found.
With a hardware deployment, we further show that the retargeted motions are physically feasible, while MPC ensures that the robots retain their capability to react to unexpected disturbances,” said the team.
