IN a groundbreaking move for industrial automation, UBTECH has successfully executed the world’s first collaborative practical training program for humanoid robots at ZEEKR’s state-of-the-art 5G Intelligent Factory.

This initiative represents a key advancement in developing a general-purpose swarm intelligence system for humanoid robots, marking a major leap from single-agent autonomy to swarm intelligence (SI), the Shenzhen company said in a press release yesterday.

SI is a form of collective intelligence that emerges from the interactive behavior of self-organized systems, particularly from natural and artificial systems composed of a large number of individuals that coordinate through decentralized control and self-organization.

The concept is inspired by the social behaviors of animals such as ants, bees, birds, and fish, which work together in large groups to solve complex problems or complete tasks without a central leader or controller.

The training program saw multiple Walker S1 humanoid robots from UBTECH work in unison across a variety of complex industrial tasks and scenarios, according to a video clip released by the company.

Dozens of Walker S1 robots were seen executing collaborative sorting, handling, and precision assembly tasks. The robots utilize cross-field pure vision-based perception technology and intelligent hybrid decision-making to optimize sorting tasks and ensure efficient collaboration.

To enable humanoid robots to effectively operate at scale in production-line environments, UBTECH said it has deployed a networked cluster that explores the emergence of advanced embodied intelligence through swarm behaviors and high-frequency interaction with physical surroundings.

Central to this achievement is its development of BrainNet, a software framework designed for humanoid robot collaboration, and the introduction of the “internet of humanoids” (IoH), a centralized control hub that serves as a blueprint for implementing swarm intelligence in both software and hardware.

The company also claimed to have developed the world’s first large reasoning multimodal model for humanoid robots, which serves as the core engine of the super brain.

This AI system, currently under development with DeepSeek-R1 deep reasoning technology, equips humanoid robots with human-like common-sense reasoning, enabling them to autonomously break down, schedule, and coordinate tasks in complex industrial workflows.

(Zhang Yu)