Beijing 51WORLD Digital Twin Technology Co., Ltd. announced the deployment of “51Claw,” an embodied AI system built on the open-source platform OpenClaw. The new solution fuses cloud-based multimodal large models with local edge computing to enable robots to perform spatial planning and natural-language interaction.
51Claw operates through a Real2Sim2Real closed loop. The platform collects multimodal spatial data, reconstructs scenes with 3D/4D Gaussian Splatting, and generates a physical-world model endowed with spatial memory. Leveraging reinforcement learning and physics engines within the Group’s simulation training system, the solution orchestrates tasks and motion training across large-scale scenarios before deploying matured agent models to real-world hardware. The system is already integrated with robotic dogs and humanoid robots and connected to Tencent WeCom to form a complete business-process flow.
Management views 51Claw as a strategic extension of core products SimOne and DataOne, allowing the Group’s digital-twin technology to perceive and operate directly in physical space. The initiative strengthens the company’s position in spatial computing and aligns with its goal of becoming the “first listed company in Physical AI,” opening potential incremental commercial opportunities.
Looking ahead, 51WORLD plans to build an integrated “embodied AI training platform” that combines virtual environments from its 51Sim system with real-world training grounds. The long-term vision is to create a scalable training-and-iteration loop that supports multiple categories of embodied intelligent robots, thereby expanding the Group’s infrastructure-service coverage in Physical AI.
The company cautions that software-and-hardware iteration and commercial implementation are subject to uncertainties such as AI industry policy shifts, technology-development paths, and market demand. Further announcements will follow if material progress occurs.
Shareholders and potential investors are advised to exercise caution when dealing in the company’s shares.