Chinese lab claims first humanoid robot control using space inference

A Chinese laboratory collaboration between GuoXing Aerospace Technology and Shanghai Jiao Tong University has successfully demonstrated the control of a humanoid robot using space-based computing. This pioneering experiment integrated an open-source AI agent (OpenClaw), orbital processing powered by Alibaba’s Qwen3 large language model, and a ground robot into a closed-loop control system. Voice commands issued on the ground were transmitted to satellites in low Earth orbit (LEO), where AI inference was performed in orbit before sending instructions back to the robot, marking the first instance of AI “thinking” hosted entirely in space. This approach enables high-performance AI control of robots and autonomous systems even in areas lacking terrestrial network coverage, such as disaster zones or remote wilderness. GuoXing Aerospace is rapidly expanding its satellite constellation to support this space-based AI infrastructure. After launching an initial cluster of 12 satellites in May, the company plans to deploy two more clusters this year and aims to build a massive constellation of 2,800 specialized satellites