OpenClaw Integrates Satellite AI, First-Time Achievement of Space Remote Control of Ground Robots

China National Xing Aerospace Technology Group and Shanghai Jiao Tong University Joint Laboratory have recently completed a milestone technological experiment: using the open-source AI agent OpenClaw as a bridging layer, they conducted space reasoning with large language models (LLMs) onboard satellites, enabling remote intelligent control of ground humanoid robots. This is the world’s first practical verification of deploying AI token calling services in space.

Experimental Technical Architecture: From Voice Commands to Robot Actions

The experiment demonstrated an end-to-end closed-loop control process, with OpenClaw playing a key bridging role:

• The operator issues voice commands, which are received and processed by OpenClaw
• OpenClaw uploads the commands to the satellite in orbit
• The satellite-mounted LLM utilizes space computing power to perform in-orbit reasoning and generate decision results
• The decision results are transmitted back to Earth and received by OpenClaw
• OpenClaw translates the commands and drives the ground humanoid robot to perform the corresponding actions

The core significance of this experiment lies in validating the technical feasibility of “supporting silicon-based intelligent agents with space computing power,” providing the world’s first real-world case of deep integration between AI agents and satellite computing capabilities.

Strategic Significance of Space Computing Power: AI Computing Infrastructure Independent of Ground Networks

When ground communication infrastructure is disrupted due to natural disasters, extreme terrain, or other factors, traditional AI systems relying on ground servers lose their computing support. Space computing power offers a distributed computing capability that can efficiently serve devices such as humanoid robots, quadruped robots, autonomous vehicles, and various drones.

In January this year, China National Xing Aerospace took the lead in uploading Alibaba’s Qwen3 large language model to its space-based computing center, achieving fully in-orbit end-to-end reasoning tasks, laying the foundation for this OpenClaw experiment.

China National Xing Aerospace’s Space Computing Roadmap: 1,000 Satellites by 2030, 2,800 by 2035

In May last year, China launched the first satellite constellation of the China National Xing Aerospace space-based computing project, consisting of 12 satellites. According to the company’s long-term plan, by 2035, it aims to establish a dedicated computing satellite network comprising 2,400 reasoning satellites and 400 training satellites, deployed in sun-synchronous, dawn-dusk, and low-inclination orbits at altitudes between 500 and 1,000 kilometers. The second and third batches of satellite clusters are expected to be deployed this year, and by 2030, the overall network will expand to 1,000 satellites.

Frequently Asked Questions

Q: What role did OpenClaw play in this experiment?
OpenClaw is an open-source AI agent that acts as a bridging layer in this experiment: it receives voice commands from the operator, uploads them to the satellite, receives the decision results after satellite reasoning, and drives the ground humanoid robot to perform actions. It is the key node connecting human commands, space computing power, and ground robots.

Q: Why use satellite computing power instead of ground servers for AI reasoning?
The core advantage of satellite computing power is its independence from ground network infrastructure. In extreme environments where ground communication systems are damaged or insufficiently covered, it can still provide high-performance AI computing support, enabling humanoid robots, autonomous vehicles, and drones to operate intelligently.

Q: What is the scale of China National Xing Aerospace’s satellite computing plan?
According to the plan, the company aims to build a large-scale satellite computing network with 2,400 reasoning satellites and 400 training satellites by 2035. The second and third batches of satellite clusters are expected to be deployed this year, and by 2030, the network will expand to 1,000 satellites.