Duration: 12 months
Critical space applications calling for more onboard autonomy—such as in-orbit servicing, autonomous landing, docking, rovers, drones, space situational awareness, and debris removal—demand technologies that prioritize low latency, power efficiency, and fault tolerance. Current solutions, utilizing non-radiation-hardened components like AMD Xilinx products, are innovative in their application but face challenges in energy efficiency, with power consumption ranging from 40 to 100 watts. This presents limitations for space missions. We propose to investigate the integration of emerging neuromorphic computing technology with advanced microprocessor designs to significantly improve machine learning inference capabilities in space, focusing on minimizing power consumption. Neuromorphic computing, as demonstrated by BrainChip's technology, represents a revolutionary approach to AI inference, capable of reducing power needs from tens of watts to just a few or even below one watt for less demanding tasks. Such a leap in energy efficiency, paired with fault-tolerant, radiation-hardened designs, has the potential to elevate European space technology to a leadership position in critical mission applications. Our aim is to assess the benefits of neuromorphic computing for space use cases by integrating an innovative intellectual property (IP) with a next-generation microprocessor design on FPGA. This exploration seeks not only to demonstrate the advantages of neuromorphic computing but also to lay the groundwork for a product that could be commercialized in the future. By capitalizing on European expertise in radiation-hardened microprocessors and ultra-low-power AI hardware accelerators, this initiative strives to address the emerging needs of space missions with cutting-edge, energy-efficient solutions.
