07,23,2026 1 views
During the 2026 World Artificial Intelligence Conference (WAIC), Topstar, which has been deeply involved in manufacturing scenarios for nearly 20 years, showcased breakthrough achievements in industrial embodied intelligence. At the conference venue, relying on the collaborative innovation ecosystem of China's first physical intelligence innovation consortium,Topstar, in collaboration with leading domestic physics engine and physics world model company Feijie Kesi, has released the Fyriverse industrial physics world modelBy directly addressing the industry shortcomings of traditional industrial simulation physics distortion and the inability to reuse simulation data on real machines, we aim to build a high-precision universal industrial simulation base to help Tuosida build a solid differentiated technical barrier in the industrial physics AI and embodied intelligence track.
On a technical level, Fyriverse is equipped with the self-developed Fysys full stack differentiable physics simulation engine, which can accurately reproduce the real physical characteristics of factory components, the stiffness, deformation, friction, load, tolerance, and other aspects of the entire production line. With just one click, it can generate a high fidelity interactive digital twin environment and produce training materials that match real working conditions in a 1:1 ratio. This makes industrial simulation no longer just "resembling the real world", but truly "serving real industrial scenarios"Industrial robotBy understanding the physical operation laws of the factory, one can accurately identify on-site working conditions and stably adapt to production variables, truly achieving understanding the world, knowing how to work, and stable implementation.
Relying on the cloud edge end integrated self evolving architecture,The Fyriverse model is deeply integrated with the massive industrial scene experience and landing data accumulated by Tuosida for many years, creating a complete technology loop of self driven iterative upgradingThe optimized AI algorithm and production process plan can be simulated and verified in a virtual physical environment to avoid the trial and error costs and risks of real machine debugging. Once they meet the standards, they can be quickly deployed to physical robots with just one click. At the same time, the process data, operational experience, and scene adaptation capabilities accumulated from the operation of the real robot will continue to flow back to the simulation system in reverse, forming a sustainable evolution mode of "high-precision physical modeling AI efficient training real machine landing iteration".
Tuosida wheeled humanoid robot Xiaotuo conducts material detection and verification on the injection molding production line
This cooperation is a key layout for Tuosida to improve its industrial embodied intelligent technology layout and connect the full chain capability of virtual and real integrationAt the same time, the model is fully compatible with the entire range of industrial robots from Tuosida, which can break away from the industry's long-term reliance on manual experience for repeated debugging, time-consuming and difficult to standardize old models, significantly reducing the threshold for equipment landing, and laying a solid foundation for the cross industry, standardized and large-scale mass landing of embodied intelligent robots.
Tuosida wheeled humanoid robot Xiaotuo performs sorting and tray verification on the injection molding production line
In the future, Tuosida will continue to use AI technology as the core driving force to create a reusable standardized landing paradigm for the industrial physics AI and embodied intelligence industries, promote the true and standardized application of industrial embodied intelligence in various industrial scenarios, and accelerate the intelligent upgrading process of the manufacturing industry.
Writing | Xu Jingyu
Review | Feijie Kesi, Tailored Intelligent Products Department
Source | Feijie Kesi
