Every integration, in one place
Native connections into the simulators, middleware, engines, analysis tools and third-party hardware humanoid teams already run. No adapters. No wrappers. Just connect.
Straight into the training loop
Native publishers and simulator plugins for the stack humanoid teams actually run. Motion arrives as usable signal, not as a file waiting for conversion.
ROS2
First-class ROS2 publisher. Pipe motion data into your perception, planning and control stack the moment it arrives.
ARM64 Linux & Windows
Native ARM64 Linux build for Linux SDK. Drop it directly onto the robot’s onboard compute, NVIDIA Jetson, custom SoC or any ARM64 Linux platform. Also available off-the-shelf for Windows.
NVIDIA Isaac Sim
Drive simulated humanoids with real human motion. Close the sim-to-real loop with validated biomechanical data as ground truth.
MuJoCo
Direct MuJoCo integration for RL policy training, imitation learning and physics-based evaluation of captured behaviors.
Coming soon
HTC Vive and Vive Ultimate Tracker
Plug HTC Vive and the Vive Ultimate Tracker straight into Xsens Humanoids for object tracking and absolute positional reference. Sub-millimetre spatial grounding for scene-synced data and contact-rich interaction tasks, layered into the same capture pipeline.
From human body to your humanoid
Retargeting turns a validated human capture into joint targets for your specific platform. Run our engine, or plug in the research framework your team already trusts.
XsensTargeter
Our own real-time retargeting engine. Optimized for live teleoperation with minimal latency. Currently supports Unitree G1 and AgiBot A2.
Westlake
Advanced cross-embodiment transfer algorithms from Westlake’s robotics lab. Handles diverse morphologies with physics-aware optimization.
GMR (Stanford)
Generalized Motion Retargeting framework from Stanford University. Broad humanoid compatibility with learned correspondence models for accurate motion transfer.
MPI (Max Planck)
SMPL/SOMA body model integration from the Max Planck Institute. Physics-informed retargeting with body shape and pose estimation.
Into the DCC pipeline
BVH and FBX land in the engines and animation suites your team already uses. Useful for review, annotation and building synthetic variation on top of measured motion.
Unreal Engine
Live streaming and recorded playback
Unity
Live streaming and recorded playback
Visual3D
Biomechanical analysis
MotionBuilder / Maya
Animation retargeting workflows
Blender
FBX and BVH import
Siemens Tecnomatix
Manufacturing process simulation
OpenSim
Musculoskeletal modeling and simulation
Dassault Delmia
Digital factory simulation
Expand the setup with best-in-class modules
Both Live Teleoperation and Human Movement Data Collection share the same optional modules. Layer the ones that match the task you actually want the humanoid to perform.
Xsens Metagloves by Manus
Full per-finger degrees of freedom inside the same capture loop. Powers dexterous manipulation, grasping and contact-rich tasks in live teleoperation or offline datasets.
StretchSense
Soft sensor gloves for finger motion capture. An alternative glove route into the same per-finger data stream, on the same capture timeline.
HTC Vive & Vive Ultimate Tracker
Track tools, payloads and workspace objects alongside the human, and lock the suit to a global reference frame. Sub-millimetre spatial grounding for scene-synced data and contact-rich interaction tasks.
NOKOV
Optical tracking systems from the Orbit and Mars series. Hybrid IMU plus optical capture for absolute positional reference layered onto the same pipeline.
Delsys Trigno
Wireless EMG layered on top of kinematic ground truth. The basis for human-in-the-loop control research and humanoid teleoperation studies.
Meta Quest & Pico VR
Pair captures with VR headset data for first-person datasets, immersive review and full-body teleoperation.
Coming soon
Multi-camera video sync
Time-locked multi-camera video streamed alongside motion. Vision and biomechanics on the same clock, ready for VLA training and annotation.
One dataset, many exports
Xsens captures once and exports into every format humanoid teams use for model training, simulation, retargeting and analysis.
BVH
Skeleton animation format. Lightweight, widely supported across animation and RL training frameworks.
FBX
Industry standard for animation tooling. Position and orientation data for MotionBuilder, Maya, Blender, Unreal and Unity.
C3D
3D coordinate format for biomechanics and robotics analysis tools. Native support in Visual3D and academic research stacks.
MVNX
Open XML format. Full segment, sensor and joint data in a human-readable schema any team can parse and drop into a custom training or analysis pipeline. Direct import into MATLAB, Excel and Visual3D.
SMPL
Parametric human body model used across humanoid imitation learning, diffusion policies and VLA training.
CSV
Tabular export for custom pipelines, PyTorch dataloaders and lightweight analytics.
XLS
Spreadsheet export for rapid inspection, joint angle review and stakeholder reporting.
URDF-ready
Retargeted joint streams matched to humanoid URDF conventions for direct use in simulators and policy training.
MJCF
MuJoCo MJCF export for RL policy training and physics-based evaluation.
ROS2
Native rosbag logging for ROS 2 and ROS 1 stacks.
MCAP
Robotics data format for NVIDIA Isaac Teleop workflows. Packages motion data for recording and replay in Isaac environments, supporting simulation, debugging and robot training pipelines.
Missing the integration you need?
Tell us which simulator, engine or hardware your pipeline runs on. We will map the right route in, or build it with you.