NAVI-Orbital Achieves First In-Orbit Zero-Shot VLM for Earth Observation

A new system called NAVI-Orbital has successfully performed the first in-orbit demonstration of a vision-language model (VLM) for autonomous Earth observation. Deployed on a Low Earth Orbit (LEO) spacecraft, this software system utilizes a local VLM (Gemma 3) to process satellite imagery directly onboard. The system can classify captured scenes, generate detailed text descriptions of their content and feature relationships, and engage in natural-language dialogue with operators. This capability allows for re-tasking the satellite using plain English prompts instead of complex command sequences, streamlining operations. By performing inference onboard, NAVI-Orbital inverts the traditional acquire-then-downlink-everything paradigm. This "semantic compression" significantly reduces the amount of data that needs to be downlinked, addressing the growing gap between data collection and actionable intelligence due to bandwidth limitations.