(Yicai) Oct. 16 -- Humanoid robots are taking center stage at this year’s China International Industry Fair in Shanghai, as Chinese and international companies link arms to build a stronger industrial ecosystem and accelerate their deployment on factory floors.

At this year’s CIIF, which ended today, a number of humanoid robots and embodied artificial intelligence solutions demonstrated their ability to carry out precision manufacturing tasks. Shanghai-based AgiBot’s Genie G2 robot, for example, can perform millimeter-level insertion and removal operations involving optical modules. It can also be deployed across an entire production line in just two to three hours.

Taipei-based Delta Electronics unveiled a dual-arm collaborative robotics solution powered by US chipmaker Nvidia’s computing platform, designed to handle non-standard, complex assembly processes such as AI server assembly.

China’s humanoid robotics technology is currently among the global front-runners, but the industry continues to face major bottlenecks such as a shortage of industrial data, insufficient computing power and an underdeveloped supply chain.

To address some of these issues, Germany’s Siemens announced that it will open five factories in China and six industrial application scenarios for on-site training and validation of embodied AI systems, while fellow German firm Bosch Rexroth has also set up a dedicated project team and opened production-line workstations to allow robotics companies to collect training data.

At the computing-power level, beyond Santa Clara-based Nvidia’s Thor platform, which has been adopted by several high-end robots, Chinese chip solutions are moving more rapidly toward commercial deployment.

D-Robotics’ Xuri S600 chip has been adapted for use by more than 20 robotics manufacturers. Its applications include industrial welding and power infrastructure inspections, with most projects having moved to mass production and shipment. Meanwhile, Axera Semiconductor has launched a high-performance edge AI chip which when paired with its self-developed embodied AI brain, enables industrial tasks such as electronic assembly and flexible cutting.

Key Bottlenecks

Data and computing power remain the two of the sector’s biggest challenges.

Industrial data is fragmented and often kept within closed systems, said Wang Jian, a scholar at the Chinese Academy of Engineering. Limited data sharing is also constraining the iterative improvement of robotics technology.

Even with high-end computing platforms, the industry continues to face a widespread shortage of computing resources, said a representative from Shenzhen-based D-Robotics.

The ongoing development and refinement of embodied AI models will sustain high demand for computing power over the long term. The complexity of industrial environments adds another challenge. Production lines often combine older equipment with newer machinery, making data alignment difficult. The industry is still working to improve the effective utilization of available data.

Industry insiders believe that humanoid robots at this stage can only perform well-defined, predictable tasks in factories. Their ability to adapt to complex and changing real-world working conditions remains the biggest hurdle to widespread commercial deployment.

Although humanoid robots have yet to be deployed at scale in factories, multinational companies are continuing to open their production lines for training and testing, while Chinese chipmakers are narrowing the technological gap.

With closer collaboration across the global industrial chain, large-scale deployment of humanoid robots in industrial settings is becoming an increasingly realistic prospect.

Editor: Kim Taylor