SPEED: Space Processing on Edge

Purpose and goal

SPEED aims to develop an energy-efficient edge AI platform for satellites and UAVs, significantly reducing data volumes and latency. The goal is to demonstrate a modular architecture combining CPU/GPU/FPGA with ASIC accelerators and to extend Bluestone for selective downlinking in near real-time. The project will verify that the technology enables faster, more accurate geospatial insights and will lay the foundation for AAC Clyd Space´s Sirius EDGE on-board processor as a future product.

Expected effects and result

SPEED is expected to deliver a laboratory demonstrator for energy-efficient edge AI, with significantly reduced data volume, lower latency, and improved performance compared to current solutions. The project enables faster and more reliable geospatial insights, increases the autonomy of Swedish and European space systems, and lays the foundation for commercialization through the Sirius EDGE-plattform, suitable for both satellites and UAVs.

Planned approach and implementation

The project will implement a sequential chain of requirements gathering, architecture design, algorithm and hardware development, and integration and testing. After establishing user and system requirements, a modular architecture will be defined, and a prototype hardware developed. Spacemetric will develop on-board algorithms, KTH ASIC accelerators, and AAC Clyde Space will build and integrate the platform. KPIs will be verified in the lab, and results will be documented and disseminated.