Orbbec Unveils Robot-Free Data Collection Hardware Platform to Help Customers Capture Real-World Demonstrations for Physical AI at Scale

Orbbec recently announced the launch of its Robot-Free Data Collection Hardware Platform for physical AI. Built around its EGO, UMI and WristCam product series, the platform provides standardized products as well as CM (contract manufacturing) and JDM (joint design manufacturing) services for physical AI model companies, robot OEMs, data collection service providers, data operations companies and other algorithm-driven enterprises. The platform is designed to help global customers build high-quality real-world interaction data collection capabilities more efficiently.

Leveraging its deep expertise in and broad product portfolio of robotics and AI vision, Orbbec is one of the few industry providers that combines advanced multi-sensor calibration and synchronization technologies, a full-stack vision product portfolio, and global-scale manufacturing and delivery capabilities.

Real-World Interaction Data Is Becoming Critical Infrastructure for Physical AI
As physical AI moves from laboratory demonstrations to real-world applications, demand for real-world data for robot model training is rising rapidly. The ability to collect high-quality data at low cost, with stability and at scale, is becoming a critical foundation for the industrialization of physical AI.

Robot-free data collection refers to capturing multimodal data from human operations and environmental interactions through head-mounted, wrist-mounted, handheld or close-range operating devices, without relying on a specific robot body as the data collection carrier.

Compared with collecting data directly from robot bodies, this approach makes it easier to scale data collection cost-effectively, diversify human demonstration actions, and provide a stronger data foundation for model training, teleoperation systems and robotic policy optimization.

In practical data collection, the industry still faces multiple engineering challenges.
First, the spatial relationships among multi-view devices are complex. Heterogeneous sensors such as RGB cameras, depth cameras, and IMUs require high-precision joint calibration of intrinsic parameters, including focal length, principal point, and distortion, as well as extrinsic parameters, such as spatial pose.

Second, sub-millisecond time synchronization and spatial coordinate alignment across multiple devices, such as EGO, UMI, and wrist-mounted cameras, remain difficult and directly affect the quality of multi-source data fusion.

Third, during close-range operations such as approaching, grasping, and holding objects, critical hand-object interaction details can be lost due to occlusion, limiting a model's ability to learn fine-grained manipulation behaviors.

Fourth, in large-scale deployment scenarios, the consistency, stability, and supply chain delivery efficiency of data collection devices are put to a rigorous test.

A Hardware Stack Designed to Cover Key Robot-Free Data Collection Needs
To address these pain points, Orbbec has launched the Robot-Free Data Collection Hardware Platform for physical AI. The platform forms a modular product layout around EGO, UMI and WristCam, covering typical scenarios such as first-person observation, wrist-level near-field observation and hand-object interaction detail capture. It can be flexibly configured according to customer requirements, including task type, collector scale and sensor combination.

Orbbec also provides CM and JDM services for its Robot-Free Data Collection Hardware Platform. For customers with existing hardware designs, Orbbec can provide professional manufacturing and volume delivery support. For customers that need to jointly define hardware solutions, Orbbec can provide joint design services covering vision module integration, calibration and synchronization solutions, manufacturing, and mass-production introduction.

Consistency at the Core: Full-Stack Capabilities for High-Quality Data Collection at Scale
Orbbec's long-term accumulation of technology and engineering capabilities in robotics and AI vision provides a strong foundation for its data collection business.

High-precision multimodal calibration for spatial consistency.
For fisheye lenses, high-distortion lenses, and heterogeneous sensors such as RGB, depth and IMU, Orbbec has mature joint calibration processes and self-developed calibration equipment. These capabilities enable efficient and accurate calculation of intrinsic parameters, including focal length and distortion coefficients, and extrinsic parameters, including spatial pose, ensuring strict 3D spatial alignment of data captured from multiple viewpoints.

Hardware-level multi-device synchronized triggering for temporal consistency.
For multi-device collaboration scenarios involving head-mounted EGO, wrist-mounted cameras, UMI and other devices, Orbbec has accumulated extensive experience in multi-camera hardware synchronization, external trigger mechanisms and master-slave timing design. These capabilities enable microsecond-level clock synchronization and precise frame-level alignment, fundamentally reducing temporal offset errors in multi-source data.

Scaled manufacturing and global delivery for product consistency.
Orbbec has successfully delivered multiple million-unit mass-production projects and has established a digital quality management system covering R&D, component selection, incoming materials, production, testing and after-sales service. Recently, Orbbec's RVMC factory in Vietnam officially broke ground. Going forward, the company's dual-factory strategy across China and Vietnam will enable efficient global capacity allocation and build a more resilient supply chain for global customers.

Technical depth and product breadth for flexible scenario customization.
Orbbec has a full-stack robotics vision product and technology matrix covering RGB, structured light, stereo vision, ToF and LiDAR. The company has also built long-term practical experience in underlying technologies such as multi-sensor fusion design, image signal processing (ISP), panoramic stitching and color tuning. This allows Orbbec to quickly provide optimized vision solution combinations for different collection environments and close-range operation requirements, without forcing customers to compromise around the limits of standardized products.

Empowering Physical AI Through Ecosystem Collaboration
"Physical AI is increasingly becoming a competition of real-world data production capabilities," said Len Zhong, Head of Product and Technical Support at Orbbec. "High-quality training data requires system-level capabilities across multi-view capture, depth perception, close-range vision, camera calibration, multi-device synchronization, and large-scale deployment. Through our Robot-Free Data Collection Hardware Platform, Orbbec aims to leverage its robotics and AI vision platform capabilities to help global customers lower the barrier to building data collection infrastructure and accelerate the production of real-world interaction data."

To turn this vision into reality, Orbbec is continuously driving its ecosystem expansion through strategic technical partnerships. As a major collaboration breakthrough, Orbbec's RGB-D EGO device is being adapted to LingBot-Depth 2.0, the latest spatial perception model from Robbyant (an embodied AI company within Ant Group). Trained on 150 million samples, this model delivers substantial upgrades in edge clarity, small object recognition, long-range depth estimation, and complex-environment robustness. Focusing on data collection scenarios, the device will be further adapted to an advanced commercial-grade model to continuously optimize missing depth completion, object edges, and spatial structures, delivering a precise, stable, and highly usable real-world data foundation for embodied AI training.

Looking ahead, Orbbec will continue to advance its product portfolio and scenario validation around the data collection needs of physical AI. The company aims to help global industry partners build robot training data collection capabilities that are replicable, deployable, and continuously upgradable.

About ORBBEC:
Orbbec Inc., founded in 2013 and now a global leader in robotics and AI vision technology, delivers full-stack 3D vision solutions based on structured light, stereo vision, Time-of-Flight (ToF), and LiDAR, with comprehensive R&D and manufacturing capabilities advancing "hand-eye-brain" integration and multi-sensor perception for the emerging era of Physical AI. Today, Orbbec serves over 5,000 clients across nearly 100 countries and regions, with a market share exceeding 70% in both China's service robotics vision market and South Korea's mobile robot 3D vision market.
Website: www.orbbec.com