Motion Retargeting

Converting a human's (or another robot's) motion into the joint motion a target robot can actually perform.

Motion retargeting originated as an animation technique for transferring a character's motion onto another character with different bone proportions. In embodied AI it means converting human-body or hand motion — from motion capture, video pose estimation, or a VR device — into joint angles a robot can execute. The difficulty is that limb lengths, joint counts, and ranges of motion differ between a human and a robot, so joint angles cannot simply be copied over. The usual approach frames it as an optimization problem: make key points such as the wrist and fingertips match the human's position or orientation as closely as possible, while respecting the robot's joint limits — essentially a constrained inverse-kinematics problem. It is a prerequisite step for teleoperation, learning from human video, and training humanoid robots' motion tracking; common open-source tools include GMR and dex-retargeting.

ExampleGMR can convert human motion-capture data from AMASS or LAFAN1 into joint trajectories for humanoid robots like the Unitree G1, runs in real time on a CPU, and is also used by TWIST for whole-body teleoperation.

Also called

Retargeting, Hand Retargeting

Related

General Motion Retargeting · dex-retargeting · Inverse Kinematics (IK) · Motion Capture · Motion Tracking · OmniRetarget

Sources

GMR: General Motion Retargeting (GitHub)
dex-retargeting (GitHub)