gsaw presentation using big data analytical techniques · •cosmiac at university of new mexico...
TRANSCRIPT
GSAW Presentation
Using Big Data Analytical TechniquesMarch, 2017
© 2017 COSMIAC. Published by The Aerospace Corporation with permission
Partners
• COSMIAC at University of New Mexico (UNM)– Mr. Nick Buonaiuto– Mr. Mark Louie– Dr. Jim Aarestad– Mr. Craig Kief
• Air Force Research Lab Space Vehicles Directorate– Mr. Robert Sivilli– Dr. Apoorva Bhopale
• Stinger Ghaffarian Technologies (SGT Inc.)– Mr. Rohit Mital– Mr. Dennis Mateik
Introduction
• COSMIAC proudly serves as a Tier-2 Research Center under the School of Engineering at the University of New Mexico.
• COSMIAC’s role is to promote aerospace innovation through the reliable and responsible use of configurable technology in military and defense systems.
• COSMIAC (Academia), SGT (Large Corporation), and AFRL (Government).• Over the past year, each of these organizations have contributed to the
development of the Data Laboratory for Satellite Analytics (DLSA) at COSMIAC.
• Have begun to create an academic capability for big data analytics with a particular focus on spacecraft support.
Previous Activities• Enterprise Ground System (EGS) – focus area for upcoming missions.• EGS Testlab – COSMIAC has two ground stations operating 24 hours a day downloading
mission critical data:– MC3– Unified S-Band and 900MHz – No more amateur band.
• Summer 2016 Satellite Data Project – Students funded – Work began:• Amazon Web Services (AWS)
– Large allocation of processing (EC2) & storage (S3) cloud resources.• Ice Cloud and land Elevation Satellite (ICESat) mission 2003 – 2009
– Served as data source for analytics experiments. – ICESat-2 currently in production at NASA Goddard SFC, scheduled to launch in 2018.
ICESat Data Project • Problem & Objectives – ICESat mission ended prematurely:
– Use current data analysis technologies to determine if the ICESat’sknown component failures could be verified.
– Experiment with both proprietary and open-source tools for more efficient processing of large data sets acquired from the Internet.
• Conclusions – objectives successful:– Summary statistics of ICESat energy levels show decrease and
degradation as satellite LIDAR instruments approached their abrupt end-- potential for machine learning predictive analytics.
– Energy signature images (LIDAR footprint on Earth surface) also shows deformation and degradation over time-- image recognition predictive potential.
Current Activities
• Web scraping & data mining– scanning Internet for satellite
missions based on TLEs
• Satellite Integration Laboratory
• Intellectual Infrastructure• #Tweet @SocialMedia• Nvidia Deep Learning• Weighted sums from
training data ported to TX1 for deep learning on a embedded platform
Current Activities
• Hardware – No small feat getting big data laboratory into a small closet:– 8 x Xeon E5-2697v3, 112 total cores– 1Tb RAM, 30Tb storage– 10Gb & Fiber network– 8’ x 10’ cell/office/foxhole.
• Intern Support– UNM School of Engineering– Chemical, computer, electrical, mechanical, and nuclear
engineering interns at work and research. Nick and Mark, Data Laboratory for Satellite Analytics (DLSA)
Future Work
• SIL – Is it possible to train future satellite designers to take advantage of big data capabilities during the satellite design phase?
• Apps – Seeking to make inputs to big data systems satellite agnostic.• Our ground station -> Testing for the EGS• Content delivery network
– Reoccurring requested information• According to Nick: Big data is no big deal; it can happen without much more startup
infrastructure than most people already have, and really should be happening as business-as-usual, not a special new thing. – Local (100% ownable) or cloud (100% scalable) or both.– All data can be big data, and ought to be analyzed.
