- Published on
X Square Robot Launches Embodied AI Data Collection Platform Quanxta Zero Series
- Authors
- Name
- DeepHorus
Chinese humanoid robot maker X Square Robot has launched Quanxta Zero, a software-and-hardware platform designed to streamline data collection, processing, and utilization for embodied AI models. The platform aims to address critical bottlenecks in robotics data, including limited volume, slow collection speed, and inconsistent quality. By integrating collection, synchronization, cleaning, annotation, model training, and evaluation into a closed-loop workflow, Quanxta Zero seeks to transform fragmented hardware tasks into an industrial-grade data production pipeline for general-purpose robots.
Traditional teleoperation systems for robots are often costly, difficult to deploy, and generate raw data that loses value during downstream training. X Square Robot explicitly designed Quanxta Zero to overcome these shortcomings, turning embodied data collection from a cumbersome task into an efficient, automated process. The platform includes three products tailored to different data-collection needs: the G1 headband-and-gripper system, the G0 VR-and-backpack system, and the E0 first-person context device.
The flagship G1 system captures visual, tactile, and audio data while enabling collection of full-body behaviors such as movement, dual-arm manipulation, and lifting. It integrates first-person views with close-range wrist perspectives to capture both the broader environment and fine manipulation details like grasping and contact. Sensor synchronization is maintained within 1 millisecond, with full frame-level video alignment that allows data to be replayed across different robot platforms.
In testing, the G1 achieved a collection speed of nearly 100 demonstrations per hour, which is 2.33 times faster than conventional teleoperation methods. The system features one-click start and stop controls and automated downstream annotation, allowing operators to focus solely on demonstrating actions. Additionally, the G1 eliminates the backpack used in some systems by using a headband and two grippers, and it includes a magnetic hot-swappable battery for continuous operation.
X Square Robot’s data pipeline handles the critical transformation of raw data into training-ready material. It encompasses four stages: synchronization, cleaning and annotation, quality control, and security. The synchronization stage achieves microsecond-level timestamp alignment, while the cleaning stage removes low-value segments such as pauses and failures. Multimodal foundation models perform action segmentation and semantic labeling, and AI confidence scores route data automatically or to human reviewers for quality assurance.
According to X Square Robot, the Quanxta Zero platform represents a shift from fragmented data collection to an integrated, industrial-grade workflow. By addressing the core problems of volume, speed, and consistency, the platform is expected to accelerate the development of general-purpose robots. As embodied AI continues to advance, such data infrastructure could become essential for training robots capable of performing diverse, real-world tasks autonomously.
