anticensor
7.5K posts
World’s first humanoid robot swarm tackles complex factory tasks at China’s car plant | Jiji Malayil, Interesting Engineering Powered by DeepSeek-R1, Walker S1 robots use human-like reasoning to autonomously plan and coordinate tasks in dynamic settings. Chinese robotics firm UBTech’s industrial humanoid, Walker S1, has completed the world’s first collaborative training involving multiple robots, tasks, and scenarios. The milestone was achieved at a 5G-wired smart factory operated by Zeekr, the electric vehicle (EV) brand under Chinese automaker Geely. In March, using UBTECH’s advanced “BrainNet” AI framework, a fleet of the firm’s Walker S1 humanoid robots was assigned to collaboratively learning and perform complex tasks at Zeekr’s state-of-the-art smart factory in China’s Zhejiang province. In January, the firm unveiled plans to mass-produce its Walker S humanoid robots, targeting 500 to 1,000 units by year-end for industrial applications. Factory intelligence unleashed UBTECH has advanced industrial automation with its BrainNet AI framework, enabling its humanoid robot, Walker S1, to collaborate effectively on complex factory tasks. Recently deployed at Zeekr’s 5G smart factory, the Walker S1 robots demonstrated the world’s first multi-robot, multi-task, and multi-scenario collaborative training in an industrial setting. BrainNet connects cloud-device inference and skill nodes to form a Swarm Intelligence system composed of a central “super brain” and distributed “intelligent sub-brains.” The super brain uses a large, multimodal reasoning model to manage complex production-line decisions, while the sub-brain, built on a Transformer model, handles multi-robot control and cross-field sensory integration. This system supports distributed learning, accelerating task training and skill transfer among robots. Developed with DeepSeek-R1 deep reasoning technology, the model processes large-scale data for human-like common-sense reasoning. This enables Walker S1 robots to schedule, decompose, and coordinate tasks autonomously, improving decision-making and efficiency in dynamic environments. The system uses Retrieval-Augmented Generation (RAG) to enhance adaptability, accuracy, and scalability across various industrial tasks. At Zeekr’s factory, dozens of Walker S1 robots operated in critical production areas, including final assembly, instrumentation, inspection, and vehicle assembly. They perform collaborative sorting, handling, and precision assembly using hybrid decision-making and vision-based perception. Their swarm intelligence integrates semantic VSLAM navigation, collective mapping, and cloud-assisted decision-making. Walker S1’s joint planning system supports collaborative handling by adapting to environments, planning efficient trajectories, and balancing loads. Robots dynamically adjust force and position for the stable movement of large objects. In precision assembly, they manipulate deformable materials with high-precision sensors and adaptive control to prevent damage. The robots combine vision-based global positioning with force-based adjustments through reinforcement learning for quality inspection, enabling precise and adaptive evaluations. ... According to UBTech, a full fleet has officially commenced operational deployment at a 5G smart factory of Zeekr. “The Walker S series is currently undergoing trials at facilities of Geely, BYD, Foxconn, and SF Express. After more than a decade of exploration, we’re entering a new phase of development, with over 500 units in the pipeline as intended orders,” said Zhou Jian, UBTECH founder, as the People’s Daily Online reported. Read more: interestingengineering.com/innovation/chi…
In China werden zunehmend vollständig automatisierte Fabriken eingesetzt, die als „Dark Factories“ bekannt sind. Diese Anlagen arbeiten rund um die Uhr ausschließlich mit Robotern und ohne Personal vor Ort. Um Energie zu sparen, bleiben die Lichter in den Hallen dauerhaft ausgeschaltet. Welche Auswirkungen dieser Trend auf die Industrieproduktion und den Arbeitsmarkt haben wird, bleibt offen.
