Special Session: Exploring the Potential of Versal ACAP: Advancing Onboard Edge AI for Spacecraft

In the space domain, the necessity for Artificial Intelligence (AI) technologies and computationally demanding applications has become increasingly evident. This need arises from the growing complexity of space missions and the corresponding objective to enhance spacecraft decision-making capabilities. Traditional approaches have faced limitations in terms of flexibility as well as computational constraints, demanding a new perspective. With a focus on addressing these challenges, this investigation delves into the potential of the Versal Adaptive Compute Acceleration Platform (ACAP) developed by Xilinx. This hardware solution promises to enable efficient data processing and real-time inference for Edge AI applications on board spacecraft. The final objective of this ongoing study is to explore Versal ACAP capabilities on a range of use cases, including AI-driven telemetry anomaly detection and Synthetic Aperture Radar (SAR) data processing. AI-based telemetry anomaly detection occupies a critical space in spacecraft systems, as it promises to improve Fault Detection, Isolation, and Recovery (FDIR) functionalities. Traditionally, SAR data processing has been executed on the ground, whereas the Versal ACAP presents the opportunity to perform preliminary SAR data analysis onboard, enabling rapid extraction of information and facilitating more agile mission planning and decision-making. While the Versal ACAP promises to advance Edge AI applications in space, it also presents relevant challenges: power consumption and thermal dissipation pose a critical concern as spacecraft systems operate within strict power budgets and limited dissipation capabilities. Striking a delicate balance between the computational requirements of AI algorithms and spacecraft design constraints becomes a crucial consideration for successful deployment. This work presents an initial investigation into the Versal ACAP through the development of a prototype board, auxiliary to traditional On Board Computers, designed to enhance on board processing capabilities. The available software (SW) tools and the proposed SW architecture required to run the application SW on the board are also discussed. Finally, the plan for the successive validation of the prototype through the use cases mentioned above is outlined.