UBTECH has successfully conducted the world’s first collaborative practical training programme for humanoid robots at ZEEKR’s 5G Intelligent Factory. This milestone marks a major leap in industrial robotics, showcasing how multiple humanoid robots can collaborate seamlessly across complex manufacturing environments.
The initiative is a key step towards developing a Swarm Intelligence system—an advanced network allowing robots to operate collectively rather than as isolated units. Traditionally, humanoid robots have worked independently, but UBTECH’s latest breakthrough enables them to function as an integrated workforce, improving efficiency in large-scale industrial applications.
Advancing Swarm Intelligence
To achieve this, UBTECH has introduced BrainNet, a cloud-based software framework enabling real-time collaboration between humanoid robots. The system operates through two components: a super brain, which uses a large reasoning multimodal model for intelligent decision-making, and an intelligent sub-brain, which facilitates multi-robot coordination and learning.
This approach allows robots to share data, analyse tasks collectively, and optimise their workflow. In practice, it means robots can distribute labour dynamically, adapting to real-time production demands and refining their movements based on shared experience.
UBTECH has also pioneered the Internet of Humanoids (IoH), a central control hub that standardises how humanoid robots interact within factory settings. Combined with BrainNet, this system enables rapid skill generation and task execution, bringing UBTECH’s humanoid robots closer to human-like adaptability in manufacturing.
Real-World Deployment at ZEEKR Factory
At ZEEKR’s 5G Intelligent Factory, dozens of UBTECH’s Walker S1 humanoid robots were deployed across key production areas, including final assembly, quality inspection, and vehicle assembly. The robots performed three core tasks: collaborative sorting, handling, and precision assembly, each requiring advanced multi-robot coordination.
In collaborative sorting, the robots used vision-based perception and shared intelligence to track and categorise components. The system’s semantic VSLAM navigation allowed them to dynamically adjust their paths, ensuring efficiency in fast-moving environments.
During collaborative handling, robots overcame traditional limitations such as uneven load distribution and trajectory planning. UBTECH’s joint planning and control system enabled them to dynamically adjust their posture and grip strength, ensuring the stable transportation of large and heavy workpieces.
In precision assembly, the robots showcased their dexterity by handling deformable materials. Using high-precision tactile sensors and adaptive control, they successfully manipulated small film components without damage—an essential capability for high-end automotive manufacturing.
The Future of Swarm Intelligence
UBTECH’s achievement at ZEEKR is a significant step towards fully autonomous, AI-driven manufacturing. The company’s large reasoning multimodal model, currently in development, is designed to further enhance humanoid robots’ ability to process data at scale and make human-like decisions.
UBTECH has already deployed its Walker S humanoid robots in leading automotive factories, including those of Geely Auto, FAW-Volkswagen, Audi FAW, BYD, and Foxconn. With its Practical Training 2.0 initiative, the company is expanding deployment to new industrial partners, accelerating the real-world adoption of Swarm Intelligence in manufacturing.
