Data & collection
Motion capture
Motion capture is the measurement and reconstruction of a person’s or object’s movement over time, commonly as joint positions, orientations or a fitted body model. Optical markers, cameras and inertial sensors can supply the measurements. Humanoid robotics uses the resulting motion sequences for analysis, imitation and retargeting to a robot body.
Also known as: mocap
Updated
How motion is captured
Optical systems track visible markers or estimate pose from images. Inertial systems use body-worn accelerometers and gyroscopes to infer segment orientation and movement. Hybrid systems combine signals to reduce weaknesses such as visual occlusion or inertial drift. The Physical Inertial Poser paper demonstrates full-body tracking from six inertial measurement units with physics-based correction.
The output may be raw marker positions, skeletal joint rotations, root motion or a fitted body mesh. AMASS converts multiple optical marker-based datasets into a common SMPL-family body representation. That processing makes sources easier to combine, but it also embeds choices about the body model, coordinate frames and pose fitting.
From human motion to humanoid data
Captured human movement is not a robot control command. A humanoid has different limb lengths, joint limits, mass distribution, actuators and feet. Motion retargeting must map the recording to that body while preserving contacts and balance as far as the task requires. Fast or acrobatic human poses may remain physically infeasible.
Motion capture is useful as a reference for locomotion, whole-body skills and natural movement priors. DeepMimic used captured clips as imitation targets while reinforcement learning produced physics-based control. For reusable training data, a motion-capture release should state its skeleton or body model, units, frame rate, coordinate convention, calibration and any contact labels. Object state, force and intent may need separate sensing or annotation.
Sources
Related terms
Simulation & transfer
Motion retargeting
Motion retargeting is the adaptation of a recorded or generated motion from one body to another with different proportions, joints or limits. For humanoid robots, it maps source poses or trajectories into robot configurations while preserving task-relevant relationships such as contacts and end-effector paths and satisfying kinematic, balance, collision and actuator constraints.
Data & collection
Trajectory
A trajectory is a time-ordered sequence of states or observations, actions and, where applicable, rewards generated as an agent or robot evolves. A complete episode or policy rollout often yields a trajectory, but the terms are not universally identical: trajectories may be partial, while episodes have dataset- or environment-defined boundaries.
Data & collection
Demonstration
A demonstration is a recorded example of how an intended task or behaviour is performed, usually represented as a time-aligned sequence of observations, states and actions. For humanoid robot learning, demonstrations may come from teleoperation, kinaesthetic guidance, motion capture or autonomous experts and provide targets for imitation.
Models & learning
Reinforcement learning
Reinforcement learning is a method in which an agent learns a policy by interacting with an environment and optimising cumulative reward. In humanoid robotics, actions change the robot and world, while observations, rewards and episode endings provide experience for improving balance, locomotion or manipulation behaviour.