#China #Shenzhen #UBTECH #Walker S1 #Swarm Intelligence

UBTECH's Innovative Collaborative Robot Training

In a groundbreaking move, UBTECH Robotics has successfully executed the world's inaugural collaborative training program for humanoid robots at ZEEKR's state-of-the-art 5G Intelligent Factory in Shenzhen, China. This innovative initiative brings together multiple humanoid robots, particularly the Walker S1 model, showcasing an unprecedented example of 'Swarm Intelligence' in a complex industrial setting.

As industrial automation continues to burgeon, the collaboration between humanoid robots presents unique challenges, prompting advancements in swarm intelligence. This approach signifies a shift from single-agent autonomy to a collective operational synergy, enabling humanoid robots to tackle multifaceted tasks across diverse environments effectively.

The Essence of Swarm Intelligence

UBTECH's humanoid robots are pioneering a networked cluster, delving into how advanced body intelligence can materialize through swarm behaviors and high-frequency interactions within their physical surroundings. Underpinning this mechanism is BrainNet – a proprietary software framework facilitating collaboration among humanoid robots. It operates alongside the Internet of Humanoids (IoH), a centralized control hub that encapsulates the architecture for both hardware and software implementation of swarm intelligence.

The BrainNet framework interlinks collaborative inference nodes operated via cloud technology with skill nodes, creating a sophisticated system that mimics a superbrain and subordinate cognitive network. Powered by a robust multimodal reasoning model, the superbrain offers smart hybrid decision-making and efficiently manages complex production line responsibilities. Complementarily, the subordinate intelligent network, modeled on transformer technology, integrates cross-domain perception and collaborative control, promoting parallel distributed learning that accelerates skill acquisition and transfer.

These technological marvels empower humanoid robots to transcend individual task execution, facilitating flexible collaboration on production lines. Such dynamics are essential in navigating the complexities of modern industrial environments, thereby laying the groundwork for the next evolution in smart manufacturing.

Enhanced Decision-Making Capabilities

To address the high-dimensional decision-making demands posed by intricate production tasks, UBTECH has created the world's first large-scale multimodal reasoning model tailored for humanoid robots. As the core of the superbrain, this AI system is designed for continuous self-evolution within BrainNet, unlocking the full potential of swarm intelligence. Presently in development via the DeepSeek-R1 deep reasoning technology, the model aims to process data at large scales and equip humanoids with human-like common sense reasoning, enabling autonomous task breakdown, programming, and coordination while optimizing inter-robot collaboration in complex workflows.

Derived from high-quality industrial datasets gathered through hands-on training with the Walker S series, UBTECH's multitask reasoning model employs augmented retrieval generation (RAG) technology to accelerate adaptation to specialized job functions. This significantly enhances decision-making accuracy, cross-station generalization, and scalability for extensive industrial implementations.

Practical Training 2.0: A New Horizon

UBTECH's humanoid robots have advanced to Practical Training Phase 2.0, marking the transition from isolated agent autonomy to cohesive swarm intelligence. At ZEEKR's intelligent factory, numerous Walker S1 robots are deployed in intricate production zones, including final assembly, SPS instrumentation, quality inspection areas, and vehicle assembly stations. Collectively, these robots successfully completed collaborative sorting, cooperative manipulation, and precision assembly, thereby exemplifying seamless teamwork across real-world industrial settings.

During collaborative sorting, UBTECH's Walker S1 robots harness vision-based perception and intelligent hybrid decision-making, optimizing sorting tasks effectively. Through continuous tracking of dynamic targets across varied environments using pure cross-field vision, the robots engage in swarm collaboration via collective mapping and shared intelligence, underscored by a hybrid decision-making system integrating VSLAM navigation and adept manipulation capabilities.

Collaboration in manipulation presents unique challenges, including uneven load distribution, complex trajectory planning, and dynamic environmental adaptation. UBTECH's solution is a joint planning and control system, facilitating collaborative trajectory planning, load recognition, and adaptive compliance control among multiple robots. This ensures dynamic posture and force adjustments during manipulation, significantly enhancing stability and efficiency in moving large, heavy materials.

During precision assembly, Walker S1 robots exhibit exceptional capabilities, particularly in handling delicate, deformable objects. By leveraging high-precision detection technologies and adaptive control, these robots dynamically adjust grip strength and posture, ensuring that fragile items remain undamaged and correctly aligned throughout assembly processes.

Moreover, Walker S1 incorporates global positioning based on vision with secondary force-based positioning through reinforcement learning to perform quality inspections with remarkable accuracy and adaptability.

Collaborations and Future Initiatives

Collaborating with key industry players such as Dongfeng Liuzhou Motor, Geely Auto, FAW-Volkswagen, Audi FAW, BYD, BAIC New Energy, Foxconn, and SF Express, UBTECH leads the way in humanoid robotics. Following the successful completion of Phase 1 training in autonomous single-agent intelligence, the release of Practical Training Phase 2.0 expedites the collaborative training of multi-humanoid robots and the development of advanced AI models, fueling product and technological innovations.

As UBTECH pushes the boundaries of humanoid robotics beyond Geely Auto, it is expanding deployments to additional partner factories, further reinforcing scalable adoption in industrial applications. With this new training paradigm, the future of humanoid robots in smart manufacturing looks bright.