RealMan Robotics launches Humanoid Robotics Data Training Center in Beijing

Embodied intelligence and modular robotic systems provider RealMan Robotics has launched a large-scale Humanoid Robotics Data Training Center in Beijing, China.

The company said the new, innovative hub unites core technology R&D, scenario-based application testing, operator training and ecosystem collaboration. As a core technology and equipment provider, RealMan Robotics said it plays a central role in both the deployment and daily operations of the center.

World-class robotics training facility

Spanning 3,000 square meters, the center is divided into a training zone and an application zone, with 108 robots of diverse forms already deployed. These include embodied dual-arm lifting robots, wheeled humanoids, drone-arms and quadruped robotic platforms.

To ensure data quality and scenario realism, RealMan said the center has constructed ten real-world environments - including eldercare and rehabilitation, special operations, new retail, automotive assembly, and smart catering. Together, Realman said these scenarios support large-scale multimodal data generation, producing an estimated over one million high-quality data points annually for training advanced AI models.

RealMan said the center addresses three fundamental pain points in robotics:

• Lack of cross-scenario data generalization
• Significant gaps between simulation and real-world conditions
• Absence of standardized data formats and efficient closed-loop iteration

By creating a full-stack data pipeline - from collection and training to validation and deployment - the center aims to accelerate the commercialization of humanoid robotics and embodied AI.

RealMan launches RealBOT Open Platform

At the center’s Open Day, Eric Zheng, director of the Humanoid Robotics Data Training Center, delivered a keynote titled “Exploring the endgame of robotics.”

“Robots face three enduring bottlenecks before they can scale into everyday life: operational capability, generalization, and cost efficiency,” Zheng said. “Traditional industrial arms are heavy and expensive, service robots remain too simplistic and most lack the adaptability of humans in complex environments. Long deployment cycles and poor scenario adaptability - combined with high costs - continue to limit adoption.”

He emphasized that solving these challenges requires both breakthroughs in robot design and large-scale real-world data generation, fueling models that enable flexible and affordable deployment.

In response to Zheng’s speech and in conjunction with the Humanoid Robotics Data Training Center, RealMan unveiled the RealBOT Embodied Intelligence Open Platform, designed for high-quality data acquisition.

By deeply integrating with remote teleoperation systems, Realman said the platform creates new paradigms of human-robot collaboration.

The company added that these developments represent a key step in robotics evolving from “reliant on humans” to “assisting humans,” and finally to “empowering and liberating humans.”

Looking forward, the training center will look to expand industry-academia collaboration, mobilize ecosystem resources, and foster a culture of technology co-creation, data sharing and business co-growth.

These efforts aim to accelerate the global adoption of humanoid robotics and promote sustainable, high-quality industry development.