AI for Earth and Space: Call for researchers and experts

Frontier Development Lab (FDL) is kicking off its 7th year with a call for applications and the search for an expanded faculty. This year will be the program’s biggest and most ambitious to date, with more teams tackling challenges in space, Earth science and energy domains.

FDL is a public-private partnership between NASA, DOE, the SETI Institute, Trillium Technologies, the European Space Agency and leaders in commercial AI, space exploration and Earth science including Google Cloud, NVIDIA, Intel, IBM and Microsoft. FDL applies advanced AI and machine learning techniques to basic research problems to push the frontiers of science and develop new tools to help solve critical challenges. FDL programs benefit our partner stakeholders and all humankind. The deadline for applications is April 3, 2022.

“FDL is a stunning example of how public/private partnerships and interdisciplinary skill sets can combine to achieve extraordinary results,” said Bill Diamond, CEO, SETI Institute. “But these results also derive from the efforts of extraordinary researchers and mentors. We’re thrilled to be launching our most ambitious FDL sprint ever this year, and invite you to join us!”

“AI is becoming a powerful partner for space exploration. AI is also proving to be a crucial ally for managing our planetary spaceship, Earth. If you’re interested in working on problems that will help define the future of our species, on and off-world, then we’d love to hear from you” said James Parr, FDL Director.

FDL tackles knowledge gaps in space science by pairing machine learning experts with domain experts. Research teams are supported by computer data and advisory from our private sector partners for an intensive eight-week, paid research sprint over the summer between June and August. Final 2022 challenges will be announced in March, but research areas include:

Lunar Exploration: We are building on FDL’s growing partnership with the Luxembourg Space Agency (LSA) to support this decade’s ambitious lunar exploration goals. Can Physics-Informed Neural Nets (PINNs) replace traditional methods in lunar mapping to support rover traverses and human operations at the lunar poles?

Space Medicine: For the first time in half a century, astronauts will soon be living and working for long durations in deep space – either on Gateway, Artemis missions to the Moon or perhaps, in the 2030s, Mars. We know that the absence of gravity and increased solar and cosmic radiation conspire to make numerous challenges to long-term mission operations. Can ML techniques such as causal inference unlock powerful tools to unlock interventions on a molecular level?

Astrobiology: Finding the ingredients for extraterrestrial life (sometimes called ‘biosignatures’) is a task well-suited to ML – whether it is scanning through vast data troves or simplifying laborious workflows. Can ML help develop better definitions of ‘life’ to support rover-based exploration or large-scale all-sky surveys?

ML Onboard: As ML-enabled processors get more power-efficient and radiation tolerant, the opportunity is opening up for more intelligence and autonomy on our spacecraft – as recently demonstrated by the TRN (Terrain Relative Navigation) pipeline running on NASA JPL’s Perseverance Mission. Can we use ML to develop smart constellations for more effective disaster response, planetary management or deep space exploration?

Climate Adaptation: Adapting our civilization’s key infrastructure to a rapidly warming climate will inevitably be at the top of our species’ “To Do” list for the coming decades. Can ML help probe geomechanical data to future-proof our energy grid or provide insights on strategies for carbon sequestration? Can ML improve our decisions on building and managing the utilities of the future?

Disaster Response: FDL has already shown how ML can support Disaster response – from rapid flood mapping and inundation warning to wildfire ignition, predicting fire-spread and lightning flash rate. However, there is much work to enable operational and trusted systems. Can ML better support disaster response from resilience planning through to just-in-time insight and post-disaster recovery?

Energy Futures: ML is proving a powerful tool for assisted discovery and has already proved to be a powerful tool in drug discovery and engineering. Can we use techniques such as NLP (Natural Language Processing), Reinforcement Learning (RL) and genetic selection algorithms to accelerate the development and management of zero-emission energy solutions?

Earth Science: Over the past decade, there has been a quiet revolution in Earth Observation (EO) technologies, with constellations of Earth-orbiting satellites now providing a daily view of our planet from multiple vantage points and a wide variety of instrumentation. Can we pair this deluge of data with ML to uncover exciting new opportunities for understanding fundamental causal processes of our planet, from atmospheric interactions to hydrology, through to the origins of drought and other indicators of the health of critical ecosystems?

Live Twin: Earth System Predictability (ESP) has emerged as a key ‘moonshot’ for simulation science and High-Performance Computing (HPC). FDL has already taken small steps towards this vision, showing the potential for ML to emulate empirical models with greater efficiency, often reaching HPC parity with a fraction of the power. Can ML further push the state-of-the-art in creating a digital twin of our planet’s systems – twinning in real-time, a so-called “Live Twin”?

Heliophysics: Our local star remains the biggest influence on our planet, and its behavior is the most significant unknown variable for deep space exploration and habitability. FDL has built a broad portfolio of ML pipelines, from predicting thermospheric drag to starspots on distant stars. Can we use ML to better understand and predict the Sun’s influence on climate and make deep space exploration safer?