Post-Launch Capability Expansion of Vision-Language Models via Prompting for On-Orbit Spacecraft Inspection
arXiv:2606.15427v1 Announce Type: cross Abstract: Spaceborne inspection systems often deploy perception models prior to launch, after which updating model weights or expanding fixed label sets becomes operationally impractical. While supervised models can be integrated pre-flight, adding new semantic capabilities in orbit requires retraining and re-uploading parameters. We investigate whether prompt-driven vision--language models can enable post-launch semantic expansion, allowing new spacecraft components to be specified via natural-language prompts without modifying onboard weights. We evalu
The increasing sophistication and miniaturization of satellite technology, coupled with advancements in Vision-Language Models (VLMs), makes in-orbit AI capability expansion a critical and timely innovation.
This research outlines a method to significantly enhance the longevity and adaptability of spaceborne AI systems, reducing the need for costly and complex hardware updates for perception models.
Spacecraft inspection and other in-orbit perception tasks can now evolve with new requirements through software updates via prompts, rather than requiring physical payload modifications or complete model re-uploads.
- · Spacecraft operators
- · Satellite manufacturers
- · AI software developers
- · Defence sectors
- · Companies relying on frequent physical payload upgrades
- · Legacy AI update methodologies
Reduced operational costs and increased adaptability for space-based assets through enhanced AI capabilities.
Faster development cycles for space missions, as AI systems can be refined and expanded post-launch based on evolving needs.
Enhanced strategic capabilities for on-orbit servicing, surveillance, and national security through more flexible and intelligent space systems.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI