In a milestone fusion of three frontier technologies, a Chinese research team has successfully demonstrated controlling ground-based humanoid robots using AI agents processed in orbit. The results, published on March 18, 2026, mark the first known integration of OpenClaw with space-based computing infrastructure.
How It Works
GuoXing Aerospace Technology, a Chengdu-based commercial aerospace firm, partnered with Shanghai Jiao Tong University to build the system. Voice commands are sent to satellites, which run Alibaba's Qwen3 large language model on onboard compute. The LLM processes the request and transmits control instructions back to Earth, where OpenClaw agents execute them on humanoid robots.
The team deployed Qwen3 to their space-based computing center in January 2026, validating what they call "the feasibility of space computing for powering silicon-based intelligent agents."
Massive Satellite Network Planned
GuoXing launched its first cluster of 12 space computing satellites in May 2025. The company plans to complete a 1,000-satellite network by 2030 and deploy 2,800 specialized satellites — 2,400 for inference and 400 for training — by 2035, spanning multiple orbital configurations.
Potential Applications
The technology could support autonomous vehicles, robotic dogs, drones, and humanoid robots in environments where ground networks are unavailable — disaster zones, remote areas, and maritime operations. The integration of OpenClaw as the agent framework means the system inherits its open-source ecosystem of plugins, multi-model support, and messaging platform integrations.
For users looking to deploy OpenClaw agents without managing infrastructure, OpenClaw Launch offers managed hosting with one-click deployment — no satellites required.