VAIAS - Validation of an AI Accelerator for Space

Purpose and goal

The project addresses a growing market for energy-efficient, fault-tolerant onboard AI computing capable of replacing ground-based processing/power-hungry accelerators ill-suited for space environments. GR801 combines Brainchip Akida IP with Frontgrade Gaisler NOELV fault-tolerant CPU, to deliver reliable inference at the edge in space. The project objectives are two key commercialization enablers: radiation validation of GR801 and AI S/W ecosystem and neuromorphic demonstrator by Rapidity Space

Expected effects and result

The project will deliver a GR801 component validated for space environments and also extensions to the AI SW eco-system. The intended impact is to increase autonomy in European space systems, energy-efficient space missions, support for emerging AI-at-the-edge applications, and potential spillover into adjacent domains such as climate monitoring and robotics. This will contribute to AI hardware and software innovations and accelerating the transition of AI edge computing from lab to orbit.

Planned approach and implementation

Frontgrade Gaisler will start with a preparatory study in WP1 on use cases and customer intelligence. The results will act as a basis for work packages WP4 and WP5 where the project will focus on the ecosystem around AI software and neuromorphic sensing/computing. WP1 will be carried out by Frontgrade Gaisler and the results will be transferred to Rapidity Space which carries out WP4 and WP5. In parallel, validation of the GR801 will be performed by Frontgrade Gaisler in WP2 and WP3.