Embodied AI Glossary中文

Inertial Motion Capture

惯性动捕Advanced

A motion-capture method that straps a set of IMUs to the body to measure each segment's rotation and reconstruct full-body pose.

Inertial motion capture attaches more than a dozen inertial measurement units (IMUs, each containing a gyroscope, accelerometer, and magnetometer) to different segments of the body, measures each segment's orientation and acceleration, and then uses software to assemble these readings onto a human skeleton to reconstruct full-body pose. Xsens's system uses 17 wireless sensors; the Chinese company Noitom (诺亦腾) makes similar products. Compared with optical motion capture, which relies on multiple cameras tracking reflective markers, inertial motion capture needs no camera setup, isn't disrupted by occlusion, and can be used outdoors, in factories, or in any other setting; it's also cheaper. The tradeoff is that it can only directly compute relative pose — global position drifts over time — so it's less accurate than optical motion capture. In embodied AI, it's commonly used for whole-body teleoperation of humanoid robots and for collecting human-motion data: captured human motion goes through motion retargeting (converting human joint angles into robot joint angles) before either driving a robot to follow in real time or being used to train a policy.

ExampleAn operator wears an Xsens motion-capture suit; their full-body motion streams in real time into ROS or MuJoCo, and after retargeting it drives a humanoid robot to perform the same motion synchronously.

Also called
IMU-Based Motion Capture, MoCap Suit
Related
Motion Capture · Optical Motion Capture · Inertial Measurement Unit · Motion Retargeting · Whole-Body Teleoperation · Xsens
Sources
Motion capture - Wikipedia
Xsens Motion Capture

See it in the full glossary →