EdgeRunner, the leader in air-gapped, on-device AI agents for military and enterprise, has been collaborating with the Space Domain Awareness Tools Applications and Processing Lab (SDA TAP Lab) as part of the Apollo Accelerator Cohort 5 to enhance satellite reconnaissance capabilities. Our solution provides real-time satellite characterization and threat assessment of images, filling a critical gap in the SDA ecosystem.

The Vision-Language Model Solution

Organizations within the US Government, including the US Space Force, invest significant operator time manually reviewing imagery for objects of interest, such as satellites. EdgeRunner's solution streamlines this process with a specialized Vision-Language Model (VLM) trained on satellite imagery, delivering advanced space object intelligence through an AI-driven multi-stage analysis pipeline. By leveraging state-of-the-art computer vision and VLMs, it autonomously detects, classifies, and characterizes resident space objects, augmenting human analysis and accelerating decision-making.

The system first applies an object detection model to incoming imagery, identifying satellites and potential objects of interest. Once detected, the object is isolated and analyzed using a specialized VLM trained to extract detailed structural and functional attributes—such as bus specifications, antenna configurations, and solar array deployment. This information is then cross-referenced against a curated database of known objects, returning possible matches along with a confidence assessment.

When an object is identified as a previously cataloged asset, its entry is dynamically updated with newly observed characteristics, enhancing long-term space domain awareness. If the object is unrecognized, it is logged as a new entity, improving future detection capabilities. Crucially, when confidence is low or when an object exhibits anomalous or potentially adversarial behavior, the system flags it for human review, ensuring that operators remain in the loop for critical assessments.

By automating large-scale imagery analysis, EdgeRunner AI significantly reduces operator workload while maintaining human oversight where it matters most. This adaptive approach optimizes situational awareness, enhances decision timelines, and improves the speed and accuracy of operational decision-making. This results in a flywheel effect—accelerating threat assessment and response cycles in support of mission-critical space operations.

Participating in the SDA TAP Lab has been an invaluable experience. Our collaboration with the TAP Lab and other participating organizations provides us with the opportunity to showcase our capabilities while contributing to the broader mission of enhancing space domain awareness and threat detection.

EdgeRunner would also like to thank Turion Space for providing access to a dataset of satellite imagery that was instrumental in allowing us to build our specialized model. We are grateful for the support and guidance provided by the SDA TAP Lab and look forward to continuing our work in this exciting field.