Frontier Development Lab (FDL)
Script error: No such module "Draft topics".
Script error: No such module "AfC topic".
| File:FDL Glowing Badge.png | |
| Public–private partnership | |
| ISIN | 🆔 |
| Industry | Applied Artificial Intelligence |
| Founded 📆 | 2016 (US); 2018 (Europe) |
| Founder 👔 | James Parr, Trillium Technologies
Jason Kessler, NASA Pierre Philippe Mathieu, ESA |
Area served 🗺️ | |
| Members | |
Number of employees | |
| 🌐 Website | www.fdl.ai (US); www.fdleurope.org (Europe) |
| 📇 Address | |
| 📞 telephone | |
The Frontier Development Lab (FDL).[1] is an applied artificial intelligence research and development initiative that operates in the realm of space and planetary stewardship. It was founded by NASA’s Office of the Chief Technologist and Trillium Technologies[2] in partnership with the SETI Institute, and GPU chip maker Nvidia as the last initiative of NASA’s Asteroid Grand Challenge[3] (2013-2016).
FDL is a public-private partnership that uses interdisciplinary research to conduct a yearly research sprint. Since 2016, FDL has expanded from an initial focus on planetary defense to include challenges in heliophysics, lunar exploration and astrobiology, exploration medicine, Earth observation, and climate and energy futures.
Other public partners include the U.S. Geological Survey, the U.S. Department of Energy, the Luxembourg Space Agency, and the Canadian Space Agency.
Private partners that have provided expertise, data, funds, and other services include Lockheed Martin, Microsoft, Kx Systems, Autodesk, KBR Wyle, MIT Portugal, HPE, ISI/Pasteur Labs, Planet, Xprize, and the Mayo Clinic.
FDL’s mission is to advance the application of machine learning technologies, data science, and high-performance computing to push the frontiers of space research by developing applied AI tools and data products for space exploration and terrestrial challenges.
Research Cycle[edit]
FDL is centered around an eight-week interdisciplinary research sprint running from June to August, during the academic calendar recess. This sprint is preceded by challenge identification, challenge definition, researcher recruitment, and team selection which begins in January each year. The annual sprint ends with FDL researchers presenting their work in a showcase as well as in select academic and industry conferences in December, such as NeurIPS[4] and the American Geophysical Union (AGU)[5].
FDL research teams are interdisciplinary and international, composed of Ph.D. and postdoctoral-level researchers as well as a faculty of domain and machine learning experts. Each research team is tasked with addressing one challenge per sprint with the support of faculty advisors, expert reviewers, partner specialists, and compute resources.
History[edit]
FDL was created when senior planetary defense scientists requested that NASA start a research fellowship for early-career researchers as part of NASA’s Asteroid Grand Challenge, and its origins are outlined in “NASA's Asteroid Grand Challenge: Strategy, Results, and Lessons Learned” published in the journal Space Policy in August 2018[6]. The concept was developed by Trillium Technologies in 2015 to include data science to better engage the private sector, taking inspiration from NASA’s L5 Society summer programs.
FDL’s inaugural research cycle ran in 2016 with a focus on machine learning for planetary defense as a partnership between NASA Ames Research Center, the SETI Institute, and Trillium Technologies, as well as Nvidia and Autodesk. The partnership structure included NASA Space Portal Office personnel on a steering team, with the SETI Institute providing administration services, and Trillium Technologies running daily activities and acting as FDL.
NASA partners of FDL (2016 - 2023) include the NASA Ames Research Center[7], NASA Space Portal Office[8], NASA Marshall SpaceFlight Center (MSFC)[9], NASA’s Planetary Defense Coordination Office (PDCO)[10], NASA SMD Heliophysics, NASA SMD Planetary Science[11], NASA SMD Earth Science[12], and NASA Goddard[13]
FDL Europe[edit]
FDL Europe[14] was established as a sister program to FDL in 2018 using the same model in partnership with the European Space Agency’s Phi Lab (ESRIN), Trillium Technologies, and the University of Oxford via its Space@Oxford program with compute resources provided by Google Cloud and Nvidia. ESA ESOC Mission Operations joined the FDL Europe partnership in 2019.
FDL Europe’s private partners include Google Cloud, Nvidia, Scan Computers, Airbus, ISI/Pasteur Labs[15], D-Orbit, Planet and Space Applications Catapult[16].
Public Research[edit]
Public Research & Data Sets[edit]
A select number of FDL results and open-sourced data sets are available publicly[17]:
- Pyrocast - Machine learning pipeline for pyrocumulonimbus (pyroCb) forecasting[18]
- Characterization of Shadowed Regions at the Lunar South Pole - HORUS (Hyper-effective nOise Removal Unet Software)[19]
- ML4FLOODS - An Ecosystem of data, models and code pipelines to tackle flooding with ML[20]
- ITI For the Sun - Intercalibration of multi-instrument heliophysics data series[21]
- Sentinel-2 Super-Resolution - Multi-spectral multi-image super resolution of Sentinel-2 images and Its Effect on Building Delineation[22]
- SDOML - A Machine Learning Dataset Prepared From the NASA Solar Dynamics Observatory Mission[23]
- Tracking the Geoeffectiveness of Solar Storms - Determining new representations of “Geoeffectiveness” using deep learning[24]
- Temperature Maps of the Solar Atmosphere - DeepEM: A deep-learning approach to differential emission measure inversion using a 1x1 Convolutional Neural Network[25]
- Super-Resolution Maps of the Solar Magnetic FieldD - State of the art deep neural networks to calibrate and super-resolve historical maps of the solar magnetic field.[26]
- Digital Twin Earth - Can we lower the cost of accurate global precipitation forecasts?[27]
- Lightning and Extreme Weather - Can we use lightning observations from GOES satellites to improve predictions of severe thunderstorms and help forecasters keep the public safe?[28]
- INARA: Deep Learning Exoplanet Atmospheric Retrieval Workflow - Possible metabolisms within extraterrestrial Environmental Substrate[29]
References[edit]
- ↑ "Frontier Development Lab USA". Frontier Development Lab USA. 2023-03-08. Retrieved 2023-06-21.
- ↑ "Trillium Tech". Trillium Tech. Retrieved 2023-06-21.
- ↑ Bonilla, Dennis (2015-03-16). "Asteroid Grand Challenge". NASA. Retrieved 2023-06-21.
- ↑ "NeurIPS 2023". nips.cc. Retrieved 2023-06-21.
- ↑ "About AGU". AGU. Retrieved 2023-06-21.
- ↑ Shekhtman, Svetlana (2019-11-15). "NASA Applying AI Technologies to Problems in Space Science". NASA. Retrieved 2023-06-21.
- ↑ Colen, Jerry (2015-03-02). "NASA's Ames Research Center". NASA. Retrieved 2023-06-21.
- ↑ Vestal, Lisa (2015-04-23). "Space Portal Office". NASA. Retrieved 2023-06-21.
- ↑ Harbaugh, Jennifer (2015-02-12). "Marshall Space Flight Center". NASA. Retrieved 2023-06-21.
- ↑ Talbert, Tricia (2015-12-21). "Planetary Defense". NASA. Retrieved 2023-06-21.
- ↑ "Planetary Science | Science Mission Directorate". science.nasa.gov. Retrieved 2023-06-21.
- ↑ "NASA Earth Science | Science Mission Directorate". science.nasa.gov. Retrieved 2023-06-21.
- ↑ Adkins, Jamie (2015-02-10). "NASA's Goddard Space Flight Center". NASA. Retrieved 2023-06-21.
- ↑ "Frontier Development Lab Europe". Frontier Development Lab Europe. 2023-03-08. Retrieved 2023-06-21.
- ↑ "Pasteur Labs & ISI". simulation.science. Retrieved 2023-06-21.
- ↑ "Satellite Applications Catapult". Satellite Applications Catapult. Retrieved 2023-06-21.
- ↑ "FDL 2022". Frontier Development Lab USA. Retrieved 2023-06-21.
- ↑ Braude, Ashwin. "PYROCAST: Machine learning pipeline for pyrocumulonimbus (pyroCb) forecasting". SpaceML. Retrieved June 21, 2023.
- ↑ Bickel, V.T. "CHARACTERIZATION OF SHADOWED REGIONS AT THE LUNAR SOUTH POLE - HORUS (Hyper-effective nOise Removal Unet Software)". SpaceML. Retrieved June 21, 2023.
- ↑ Mateo-Garcia, Gonzalo. "ML4Floods: An Ecosystem of data, models and code pipelines to tackle flooding with ML". SpaceML. Retrieved June 21, 2023.
- ↑ Jarolim, R. "ITI For the Sun: Intercalibration of multi-instrument heliophysics data series". SpaceML. Retrieved June 21, 2023.
- ↑ Razzak, Muhammed. "SENTINEL-2 SUPER-RESOLUTION: Multi-spectral multi-image super resolution of Sentinel-2 images and Its Effect on Building Delineation". SpaceML. Retrieved June 21, 2023.
- ↑ Galvez, Richard. "SDOML: A Machine Learning Dataset Prepared From the NASA Solar Dynamics Observatory Mission". SpaceML. Retrieved June 21, 2023.
- ↑ Upendran, Vishal. "TRACKING THE GEOEFFECTIVENESS OF SOLAR STORMS: Determining new representations of "Geoeffectiveness" using deep learning". SpaceML. Retrieved June 21, 2023.
- ↑ Wright, Paul J. "TEMPERATURE MAPS OF THE SOLAR ATMOSPHERE: DeepEM: A deep-learning approach to differential emission measure inversion using a 1x1 Convolutional Neural Network". SpaceML. Retrieved June 21, 2023.
- ↑ Gitiaux, Xavier. "SUPER-RESOLUTION MAPS OF THE SOLAR MAGNETIC FIELD: State of the art deep neural networks to calibrate and super-resolve historical maps of the solar magnetic field". SpaceML. Retrieved June 21, 2023.
- ↑ Zantedeschi, Valentina. "DIGITAL TWIN EARTH: Can we lower the cost of accurate global precipitation forecasts?". SpaceML. Retrieved June 21, 2023.
- ↑ Vendzor-Cárdenas, Iván. "LIGHTNING AND EXTREME WEATHER: Can we use lightning observations from GOES satellites to improve predictions of severe thunderstorms and help forecasters keep the public safe?". SpaceML. Retrieved June 21, 2023.
- ↑ Himes, Michael. "INARA: DEEP LEARNING EXOPLANET ATMOSPHERIC RETRIEVAL WORKFLOW: Possible metabolisms within extraterrestrial Environmental Substrates". SpaceML. Retrieved June 21, 2023.
This article "Frontier Development Lab" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Frontier Development Lab. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
