MoCap for physical AI — humanoid training in Chemnitz
Precise AT7-50 motion capture for humanoid robotics and physical AI — synchronized force and motion data with Dynamicus, Kistler sensors, and alaska simulation (ICM Chemnitz & ART).
Physical AI needs precise reference data
Training and validating humanoid robots and physical AI workflows depend on motion and force data captured in sync. In Chemnitz, a shared setup combines motion capture, force measurement, and biomechanical simulation — in collaboration between ICM Chemnitz (Norman Hofman), ART, and the Dynamicus ecosystem.
Precise tracking with AT7-50
AT7-50 cameras provide highly accurate optical motion capture — the basis for joint angles, body poses, and integrating external measurement chains into one coordinate system.
Force measurement and Dynamicus/Recorder
The project uses Kistler 3D force sensors; force plates are also supported. All data streams are displayed online, recorded, and prepared for further calculation with Dynamicus/Recorder.
A key requirement is synchronization of force data with mocap — without a consistent time base, streams cannot be transferred reliably into simulation and training datasets.
Alignment and simulation
Sensors are aligned to the mocap coordinate system using the ART measurement tool. Action forces are computed in the multibody simulation software alaska with the Dynamicus human model.
Applications
Humanoid robotics & physical AI: reference motion and forces for data-driven learning and validation
Biomechanics & HRI: traceable human–machine data instead of isolated single-sensor measurements
ART provides precise optical tracking, metrological linkage via the measurement tool, and the stability multi-sensor pipelines need in research and physical AI — from lab instrumentation to reproducible simulation data.
