Chinese robotics firm UBTech has made a groundbreaking advancement in humanoid robot collaboration, successfully deploying multiple robots to work together in real-world industrial settings. The company’s Walker S1 robots, powered by its proprietary BrainNet framework, have demonstrated seamless teamwork at Zeekr’s state-of-the-art 5G-enabled factory.
UBTech’s cutting-edge BrainNet framework enables its humanoid robots to function as coordinated units rather than independent entities. This system connects cloud-based inference nodes and skill nodes, creating a dual-layer intelligence structure—a “super brain” handling high-level decision-making and an “intelligent sub-brain” managing robot coordination and learning.
According to the company, “Powered by UBTech’s revolutionary framework, ‘BrainNet,’ a team of Walker S1 humanoid robots works together to master complex tasks at Zeekr’s Smart Factory.”
BrainNet allows these robots to perform complex production-line tasks efficiently by leveraging multimodal reasoning models trained on real-world industrial data. The system relies on DeepSeek-R1 deep reasoning technology to process large-scale data, equipping the robots with human-like problem-solving abilities. This enables dynamic task allocation, real-time decision-making, and optimized collaboration.
UBTech has ambitious plans to mass-produce its Walker S humanoid robots, aiming for 500 to 1,000 units by the end of the year. These robots have transitioned from single-agent autonomy to advanced Swarm Intelligence, allowing for synchronized operations in key production areas such as assembly, quality inspection, and instrumentation at Zeekr’s factory.
In collaborative sorting, Walker S1 robots utilize hybrid decision-making and pure vision-based perception to monitor dynamic targets and optimize workflows. Their swarm coordination is reinforced through collective mapping and shared intelligence, enhancing precision in handling and sorting tasks.
UBTech’s robots are equipped with high-precision sensors and adaptive control systems that allow them to handle delicate materials without causing damage. Their agile robotic hands improve both flexibility and reliability in industrial applications.
“By integrating multimodal features and leveraging Retrieval-Augmented Generation (RAG) technology, the model adapts rapidly to specialized job functions, significantly improving decision-making accuracy, generalization across various workstations, and scalability for large-scale industrial deployment,” UBTech stated.
These robots also integrate vision-based global positioning with force-based secondary positioning using reinforcement learning. This technology ensures high-precision quality inspections, adapting to different tasks while maintaining efficiency and accuracy.
