Shihab-l: A Cost-Effective Spacecraft On-board Computer with Machine Learning Capabilities

Space missions classified as high mission class require fully radiation-hardened (rad-hard) electronics. Those processors are lagging the modern technology by 10–20 years in terms of performance and size. With the recent advances in Machine Learning (ML) and Computer Vision (CV) algorithms there is a need to deploy them on-board spacecraft for increased autonomy. But these algorithms require orders of magnitude more compute power than legacy processors can deliver. At the same time ML-capable spacecraft processors need to satisfy the usual strict size, weight, and power (SWaP) constraints as well as be resilient to radiation effects of outer space. Consequently, alternative onboard computer architectures are emerging. In this paper, we introduce a miniaturized <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(110 \ \text{mm x}\ 120\ \text{mm})$</tex> onboard computer designed to meet the requirements of Low Earth Orbit (LEO) and short Lunar missions, referred to as Shihab-l. The Shihab-l weighs 150 grams and is designed to operate within 5 watts power. It combines the control-flow capabilities and floating-point performance of traditional processors with the hardware-acceleration of FPGAs. This architecture features Microchip Polarfire SoC that combines 64-bit RISC-V CPUs with SEU-resilient non-flash based FPGA fabric. The Shihab-l architecture leverages both radiation-hardened (Rad-Hard) and commercial-off-the-shelf (COTS) components in a cost-effective way while ensuring there are no single points of failure (SFP). We propose to deploy compute-intensive ML and CV algorithms to FPGA fabric and keep general control and FDIR logic running on CPUs. We plan for extensive program of qualifying Shihab-l for space missions including thermal-vacuum chamber testing and radiation testing to make it available for missions starting from 2027. Our proposed solution addresses the critical need for robust and efficient computing in space, providing a pathway for more complex and autonomous missions. Additionally, the Shihab-1 onboard computer offers scalable performance, making it adaptable for various mission profiles and extending its utility across multiple space exploration projects.