Invited talk - 20 min
in
Workshop: AI for Earth and Space Science

Model Interpretability as Key Trust Element for Onboard Science Autonomy

Lukas Mandrake

2022 Invited talk - 20 min
in
Workshop: AI for Earth and Space Science

Abstract

The extremely limited communications bandwidth between Earth and distant spacecraft forms one of the greatest challenges to planetary science advancement. Rather than the 100’s of TB’s of remote sensing data now common in terrestrial remote sensing, a mission to distant ocean worlds like Enceladus or Europa may have only 75 MB of total downlink for both science observations and engineering data. Onboard science capabilities, a unique new form of autonomy, can mitigate much of this bottleneck by recognizing, summarizing, and prioritizing science observations based on their utility to ground science teams. As this requires building a new level of trust with mission scientists, these onboard systems must be constructed with reconfigurability and interpretability as primary requirements, as well as the capability to provide overlapping lines of evidence for any drawn conclusions, all with very limited onboard compute power. As these systems often incorporate machine learning and other data-driven solutions, they form a unique challenge area advancing the definition and boundaries of model interpretability for scientific insight generation.

Speaker

Lukas Mandrake

Dr. Lukas Mandrake is a senior researcher and group supervisor of the Machine Learning (ML) and Instrument Autonomy group at JPL since 2002. His group brings ML techniques to bear against a wide range of applied problems in space, science, technology, and medicine. Our products have empowered flight missions (MSL Curiosity rover, MER Opportunity/Spirit rovers, EO-1, OCO-2/3…), advanced the state of the art for commercial customers like Kaiser, Statoil, and Chevron, were infused into numerous university settings, and addressed governmental agency needs (USDA, DOE, and DOD). Specifically, Mandrake specializes on scoping a question to make it answerable given available data, followed by creating ML-based solutions that are explainable and scientifically defensible (e.g. no “black box”). Lukas is also an ML liaison to the Information and Data Science program office at JPL, bringing a “trench-view” perspective on data science to the highest levels of strategy. He is passionate about bringing Data Science and especially ML to the JPL science community as well as the Ops and On-board autonomy systems. Oftentimes, these low-hanging fruits remain untouched due to cultural and perception differences between siloed areas of specialization; Lukas has been functioning as a liaison to ease the technological transfer of ML into these high-impact areas with significant success and over twenty major publications in applied ML application.

Video

Chat is not available.