ASICS – Adaptable, Scalable, Intelligent Computing for Space

As we embark on increasingly ambitious missions in Earth orbit, to the Moon, and beyond, the need for advanced artificial intelligence and machine learning (AI/ML) capabilities becomes ever more apparent. While AI/ML technologies are often deployed on Earth using either massive cloud computing resources or specialized commercial-grade chipsets, applications for space are far more constrained. Astrobotic proposes the development of a new rugged, radiation-tolerant computing solution using a processor tailored specifically for Size, Weight, and Power (SWaP) constrained AI/ML applications: the AMD Versal Adaptive System on Chip (ASoC) Edge Series. The Versal Edge Series is a new space-grade processor with dedicated on-chip AI Engine resources that are optimized for power efficient, high bandwidth AI/ML computations. SWaP are crucial parameters for maximizing the efficiency, reliability, and performance of any spacecraft, and to date these constraints have been especially limiting in AI/ML applications requiring high computational resources. Our Phase 1 work titled Adaptable, Scalable, Intelligent Computing for Space (ASICS) will address these limitations and provide a path for artificial intelligence beyond Earth’s atmosphere via: 1. Analyzing and maturing a Versal Edge Single Board Computer (SBC) design to support board manufacturing in Phase II. A test plan will also be developed for Phase II mission specific environmental and reliability testing. 2. Developing a robust AI/ML application roadmap for the Versal Edge space computer. In support of this application roadmap, a software workflow will be developed to show how new AI/ML models will be deployed to the Versal Edge space computer. 3. Deploying an existing AI/ML model for lunar landing hazard detection to the Versal Edge VEK280 development board. 4. Measuring performance and power during hardware-in-the-loop (HiTL) testing to assess the real-world benefits of the unique Versal Edge ASoC architecture.