XERO Humanoid Trial Version, Pre-Registration Now Open

6 days ago

2 min read

XERO Humanoid Trial Version, Pre-Registration Now Open

There are many ways to train a humanoid robot, but you can't simply hand it raw human motion data. Humans and robots differ in joint structure, body proportions, and range of motion. XERO Humanoid is the software that automatically corrects for these differences, turning captured human motion into data a robot can use for training right away.

How XERO Humanoid Works

1. Capture Joint Coordinates with Optical Motion Capture

Marker-based optical motion capture records the 3D joint coordinates of human movement, frame by frame. It precisely captures the motions robots actually need — walking, grasping, balancing.

2. Automatically Retarget to the Humanoid's Joint Structure

The recorded skeleton is automatically mapped to the robot's degrees of freedom (DoF) and range of motion, correcting for differences in body proportions between human and robot. This retargeting step is the core of XERO Humanoid — rather than copying human motion directly, it recalculates the movement to fit the robot's physical structure, producing natural motion.

3. Output Directly into the Robot Training Pipeline

Once retargeted, the motion data is output directly in imitation learning and reinforcement learning pipeline formats. No additional preprocessing is required, which significantly cuts down data preparation time.

One of the most time-consuming parts of humanoid robot research and development is producing motion data for training. Manually converting human motion to fit a robot's joint structure is tedious, and often has to be redone from scratch for every robot with a different joint configuration. XERO Humanoid automates this process, turning motion capture data into something ready for robot training.