At the AI showcase at Google Cloud HQ today, with NASA, SETI and the Frontier Development Lab. I am adding notes here live, covering:
• Flood detection, prevention and response
• Lunar resource mapping with super-resolution for metal meteorites
• Enhanced predictability of GPS disturbances
• Generation of simulated biosensor data for astronaut health
• Expanding the capabilities of NASA’s Solar Dynamics Observatory
► James Parr, FDL Exec Producer
“The secret sauce of AI: space data sets are huge. Sometimes we have to move them with a plane.”
► Eugene Tu, NASA Ames Director:
“Most of the challenging problems of the future will require multidisciplinary solutions. ALL of the future NASA missions will require partnerships. Because our future missions are so multidisciplinary, we will have to partner for a much broader range of capabilities.
Artemis – return to the moon by 2024:
Why go back to the moon? It’s not like Apollo for two reasons:
1) Sustainable presence
2) Commercial and international partnerships
The only way it is sustainable is if it engages a broader space economy. We want to go to learn how to live there, and then Mars, and maybe beyond.
Dragonfly is our next billion dollar mission. We are going to Titan to look for pre-biotic chemistry. We will put a nuclear powered quad-copter on Titan. I love saying that.
2-hour entry interface. We need AI. Most communications and control can’t come from Earth.”
► Bill Diamond, CEO of SETI
Referencing the NASA Ames Director: “I love that his initials are E.T.”
► Scott Penberthy, Director of Applied AI, Google Cloud:
“If you think about a million people on Mars, you have to practice. And the logical place to practice is the moon. Within the decade, we will return to the moon on a regular basis as scientists.”
► Anna Patterson, MD of Google Gradient Ventures:
“Ray Kurzweil points out that AI will not stay on Earth. It will expands outward at the speed of light and spread through the universe. So, AI is not only the most important technology on earth, it is the most important technology in the universe.”
► MOON FOR GOOD — LUNAR RESOURCE MAPPING
Looking for metal meteorites on moon for ISRU (in-situ resource utilization). It is estimated that billions of tons of metallic deposits could exist on the Moon from M-class impactors. How might we use data fusion and emerging super-resolution techniques to develop high-resolution lunar resource maps of these metallic deposits to aid mission planners looking to locate resources for future robotic and human lunar missions? Consider Tycho. It’s deeper than Mt Whitney is tall. Optical, thermal and magnetic signatures.
(I liked this one. Seems like a much better place to mine metal asteroids — for local lunar applications and a rail gun to Earth orbit. And they have a great team)
► DISASTER PREVENTION, PROGRESS AND RESPONSE (FLOODS)
Floods are the most common natural disaster and growing. Can AI improve our capabilities to forecast and respond to floods using orbital imagery to better predict the permeability of surfaces, the likelihood of flash flooding or a burst river? Can ML techniques coupled with USGS ground observations and social data be used to better understand how to save lives in terms of better predictive models before and during a flooding event?
Frequent visible satellite imagery (Planet) enhanced with SAR (synthetic aperture radar) imagery. Made progress on real-time river level predictions. Now using ML to calibrate flood inundation maps to do automated mapping. Goal: better situational awareness.
► LIVING WITH OUR STAR: ENHANCED PREDICTABILITY OF GPS DISTURBANCES
Can we better predict atmospheric scintillations that negatively affect GPS? GPS plays a significant role in modern communication, navigation, positioning, and timing systems. GPS signals often get disrupted while passing through earth’s ionosphere, which is a volatile region of charged particles continuously getting affected by solar storms. Small-scale irregularities developed in the ionosphere as a result of solar disturbances are responsible for GPS signal disruptions, and incredibly challenging to predict at a given location and time. This challenge is to use the insight about what affects the ionospheric behavior, from the sun to the magnetosphere to aurora borealis, combined with machine learning approaches to predict GPS signal disruptions at high-latitudes.
Auroras correlate with interrupted GPS, but aurora images are diffuse and hard to analyze. Discrete structures and arcs in the aurora were found to correlate with the GPS disturbances. Several CNN layers and predictive model, but with too many false positives still. Getting better at predicting in space and time where these disturbances will occur.
Leave a Reply