Master thesis defence.
Supervisor: Jan Bergman.
Abstract:
With the growth of the space industry, we are generating more data than we can handle; creating the so-called space bottleneck. Smarter ways to filter data became crucial to space operations. In this project we have been working on designing and building a system that utilizes edge AI and vision language models (VLMs) to enable remote sensing satellites to understand not just images, but their context. The result is a system that brings world knowledge directly into orbit.
With the support of Unibap Space Solutions AB, this system was tested on next-gen space hardware; enabling the end user to communicate in natural language with the satellite and be able to simply ask questions such as "Give me a status update on the traffic situation in Sweden", "Are there any forest fires in the region" and get meaningful answers alongside filtered and relevant data, all in real time!