SPEED: Space Processing on Edge

The goal is to reduce downlink data volume and latency for satellite operators while enabling near-real-time, product-level outputs for mission end-users. SPEED addresses these needs by integrating CPU/GPU hosts with FPGA acceleration and evaluating a future path toward ASICs using KTH’s SiLago framework.

The study will implement a flat-sat demonstrator and benchmark image pre-processing algorithms using Spacemetric’s Bluestone software. Key performance indicators include latency, energy per inference, throughput, and accuracy. The results will support the development of the Sirius EDGE Platform (SEP)—a modular, scalable solution for onboard edge-AI processing in satellites and UAVs. This approach enables more autonomous, responsive space systems and reduces reliance on ground infrastructure, aligning with European goals for strategic autonomy and efficient space data exploitation.