X Square Robot has open-sourced XRZero-G0, a hardware-software framework for collecting robot training data without robots — and released the G0-Dataset, a 2,000-hour multimodal repository spanning 3,000 manipulation tasks. The Shenzhen-based embodied AI company says the system reduces real-robot training data requirements by up to 20x under experimental conditions, a direct attack on the data bottleneck slowing embodied AI.
Robot-Free Data Collection
The framework pairs an ergonomic wearable VR interface with multi-view cameras and specialized dual grippers to decouple human demonstrations from robot kinematics. A high-precision PICO 4 VR headset provides inside-out spatial tracking, while a head-mounted camera and dual wrist cameras capture both global context and detailed hand-object interactions. The system supports millimeter-accurate 6-DoF pose estimation and edge-side synchronization of visual, language, and trajectory data.
Governing Data Quality
Data quality has been the critical barrier in robot-free learning. XRZero-G0 formalizes what X Square calls trainability governance through a closed-loop collection-inspection-training-evaluation pipeline: multi-view geometric consistency suppresses visual-kinematic misalignment, full-body inverse kinematics filters invalid trajectories, and real-robot playback serves as final validation. In controlled experiments, roughly ten robot-free episodes combined with a single real-robot episode matched the performance of purely real-robot datasets.
Open Resources for Embodied AI
The code is MIT-licensed on GitHub, the dataset lives on Hugging Face, and the research paper is on arXiv — giving the community hardware designs, automated inspection pipelines, and training methodologies in one package. The release extends a run of data-centric moves in robot learning, following General Intuition's $320M raise to train robots on video game data and Acumino's seed round for robot foundation models. It also caps a big month for X Square, which just closed four consecutive funding rounds at a $2.8 billion valuation.
Reporting based on coverage from The Robot Report and X Square Robot.