Future Paths for Machine-Learning and Artificial-Intelligence in Spaceborne Microwave Instruments

The acceleration to incorporate Machine-Learning (ML) and Artificial-Intelligence (AI) into future spaceborne microwave instruments is both an exciting and challenging prospect already underway. While NASA’s Earth Science Technology Office Observation Technologies portfolio (Advanced Component Technologies and Instrument Incubation Program) does not specifically request software development, proposers have been asked since 2019 to show, where appropriate, how innovations in AI, ML, onboard processing etc, could augment instrument architecture and/or be used in the initial stages of component or subsystem design. The IIP-19 Smart ICE cloud Sensing (SMICES) instrument, which is a collaboration between JPL and Northrop Grumman, uses onboard software to detect ice storms using radiometer data and then intelligently targets storms of interest using its radar to increase science returns. The ACT-20 Stacked Miniaturized and Radiation Tolerant Intelligent Electronics (SMARTIE) project from Irvine Sensors will provide a SmallSat solution that can execute advanced AI algorithms and intensive scientific computing with a performance goal of 7 teraflops of AI engine compute performance in a 10×10 cm² SMARTIE board. This talk will highlight these two recent projects that utilize AI and ML, discuss successes and challenges experienced during development, and provide additional insights into future pathways for AI and ML in future spaceborne microwave instruments.