UBTECH Advances Humanoid Robotics with Swarm Intelligence Training at Zeekr

Shenzhen-based UBTECH Robotics has successfully completed what is claimed to be the world’s first multi-humanoid robot collaborative training program at Zeekr’s 5G Intelligent Factory. 

This marks a significant advancement in industrial automation, demonstrating the application of swarm intelligence for humanoid robots in multi-task, multi-scenario environments.

The initiative involved UBTECH’s Walker S1 humanoid robots working collaboratively across various production zones, including assembly workshops and quality inspection areas, as the company announced on LinkedIn. 

The robots executed tasks such as sorting, handling, and precision assembly, showcasing their ability to coordinate seamlessly in real-world industrial settings. UBTECH announced that this development transitions its robots from single-agent autonomy to swarm intelligence.

Advancing Swarm Intelligence in Robotics

UBTECH has developed BrainNet, a software framework enabling humanoid robots to collaborate effectively. It integrates cloud-device inference nodes and skill nodes, forming a centralised “super brain” for complex decision-making and an “intelligent sub-brain” for distributed control. 

Powered by a large reasoning multimodal model, the system allows robots to autonomously schedule and coordinate tasks.

According to UBTECH, this innovation is supported by the Internet of Humanoids (IoH), which acts as a control hub. It enables robots to adapt to dynamic environments, optimise workflows, and improve task execution accuracy.

Practical Training 2.0: Multi-Robot Collaboration

At Zeekr’s factory, dozens of Walker S1 robots participated in training that included activities like collaborative sorting, handling and precision assembly. 

Sorting uses vision-based perception and hybrid decision-making for efficient dynamic task allocation, while handling challenges like uneven load distribution through advanced path planning and adaptable control.

Additionally, precision assembly employs high-precision sensing to manipulate deformable materials without damage.

These capabilities are underpinned by UBTECH’s multimodal reasoning model, which is trained on industrial datasets from automotive factories. This model leverages retrieval-augmented generation (RAG) technology to enhance decision-making and scalability.

Expanding Industrial Applications

UBTECH collaborates with leading automakers such as Geely Auto, BYD, and Audi FAW to deploy its Walker S series robots globally. The company plans to expand its Practical Training 2.0 program to more partner factories, accelerating the adoption of humanoid robots in intelligent manufacturing.

“Swarm Intelligence represents the next frontier in robotics,” UBTECH stated. “Our innovations pave the way for scalable deployment in complex industrial workflows.”