Poster Gallery for Session 3 (February 11th): Mission Operations, Engineering, and Cross-Cutting Capabilities

Moderator: Ed McLarney, LaRC

📑 Combined PDF of posters for this session

Poster 22 Thumbnail

Time Series Analysis Methods for On-board Detection of Magnetic Field Boundaries by Europa Clipper

Authors
Gary Doran, Ameya Daigavane
Theme
Autonomy
Poster ID
22

Abstract: The Plasma Instrument for Magnetic Sounding (PIMS) on the Europa Clipper mission aims to characterize the properties of the Jovian plasma surrounding Europa, providing insight into Europa’s cryovolcanic activity and its subsurface ocean.

Poster 34 Thumbnail

GPU Saturation Testing with Variable Applications and Storage Platforms

Authors
Brian Cox, Aaron Knister
Theme
Cross-Cutting
Poster ID
34

Abstract: By design, the GPU architecture provides facilities for massive processing concurrency. Some GPU based applications distribute over 96 nodes simultaneously, touching 1500 GPUs. These application profiles necessitate parallel data paths that deliver data with high-throughput, low-latency and massive concurrency, directly to GPU memory.

Poster 35 Thumbnail

Supporting Global Knowledge Sharing using Cross-Language Information Retrieval

Authors
Petra Galuščáková and Douglas W. Oard
Theme
Cross-Cutting
Poster ID
35

Abstract: By design, the GPU architecture provides facilities for massive processing concurrency. Some GPU based applications distribute over 96 nodes simultaneously, touching 1500 GPUs. These application profiles necessitate parallel data paths that deliver data with high-throughput, low-latency and massive concurrency, directly to GPU memory.

Poster 37 Thumbnail

Solar Flare Prediction using Convolutional Neural Nets

Authors
Kendall Johnson
Theme
Cross-Cutting
Poster ID
37

Abstract: During my time with the Space Weather Lab at George Mason University (GMU), most of our research was focused on Active Regions (AR) on the Sun's surface. Recent work with Goddard's Heliophysics Lab has opened my field to the uses of Artificial Intelligence (AI) and Artificial Neural Networks (ANN). ARs are one of the last natural phenomena that we don't fully understand what governs its movements and actions. This problem was a great fit to use an ANN algorithm to determine and decipher the qualities of the images that indicate activity when formulas and simulations fail. Knowledge of the Sun's surface and ARs are critical because, at any moment, a harmful Coronal Mass Ejection (CME) can be released causing worldwide failure of the electric grid. Fortunately, most events correlate, so when a strong solar flare occurs in an active region, it is an excellent indicator that a CME will have a stronger possibility to release from that same region. Dr. Jie Zhang, a solar physic professor and advisor at GMU, and I have recently looked at the old question of can we predict solar flares from magnetogram images of the ARs using AI? We decided that using an ANN was the most efficient approach in the fact we would be dealing with larger datasets. We attempted to train the ANN with the AR images so that when the trained ANN is presented with unknown AR images, it could correctly predict if that region will have a solar flare within 24 hours. In a combined effort with GMU's computer science department, we have now matured our ANN to a Convolution Neural Network (CNN) that is optimized for image classification. CNN is still an ANN, but it has the added feature of convolution layers that mathematical takes into account the surrounding pixels as a feature of the ANN. Convolutional layers are an excellent technique used to find structures in images using only pixel data. Our research data is the magnetogram images from Helioseismic and Magnetic Imager (HMI) on the Solar Dynamic Observatory (SDO) sliced to a square region containing the full AR. Our data is from 2010 to 2014, which consists of around 1000 images. The images are from the last solar maximum to get a more significant distribution of ARs that erupted with a solar flare within 24 hrs, and this was done by connecting them with archived flare. We are now looking toward using object detection algorithms like YOLO (you only look once) to take the entire magnetogram image of Sun to detect ARs and automatically slice them to a shape the CNN can read and predict. Our end goal is the addition of these two powerful AI techniques to produce a program that can be used by scientists and satellites to predict the release of a CME on behalf of humanity. I hope to present a proof of concept that can be used to observe the Sun's surface, and when an AR forms, the object detector will find it, and the CNN determines if a solar flare will occur within 24 hrs.

Poster 38 Thumbnail

Automated Data Accountability for the Mars Science Laboratory

Authors
Brian Kahovec, Ryan Alimo, Dariush Divsalar
Theme
Cross-Cutting
Poster ID
38

Abstract: Data Accountability is the process of ensuring that all data sent from a spacecraft is received and processed successfully on the ground and also identifying where in the pipeline data becomes missing if it not received. The Mars Science Laboratory (MSL) currently relies on Ground Data Systems Analysts (GDSA) to determine whether or not all data has been accounted for. When data is missing, it can take several hours to determine the root cause of the issue. There have been previous attempts to automate the data accountability process, but they are unreliable for operational use. This paper presents machine learning based approaches to automate and optimize the detection of volume loss from the downlink process of telemetry data from the Mars Curiosity Rover.

Poster 39 Thumbnail

A Systems Engineer's Virtual Assistant (SEVA)

Authors
Jitin Krishnan, Patrick Coronado
Theme
Mission Ops
Poster ID
39

Abstract: Our goal is to develop a virtual assistant system to help and interact with one engineer in their daily lives, while gradually accumulating that specific engineer’s years of implicit knowledge and experience (lessons learned).

Poster 40 Thumbnail

AI and Data Science Using NASA’s Solar System Treks

Authors
Emily Law
Theme
Cross-Cutting and Others
Poster ID
40

Abstract: NASA’s Solar System Treks program of lunar and planetary mapping and modeling produces a suite of interactive visualization and AI/data science analysis tools. These tools enable mission planners, planetary scientists, and engineers to access mapped data products derived from big data returned from a wide range of instruments aboard a variety of past and current missions, for a growing number of planetary bodies.

Poster 41 Thumbnail

Data Ordering Genetic Optimization (DOGO) — a Data-Driven Quality Estimate for Every Observation

Authors
Lukas Mandrake, Masha Liukis, Steven Lu, James Montgomery
Theme
Cross-Cutting and Others
Poster ID
41

Abstract: (Not available.)

Poster 42 Thumbnail

Content-based Classification of Mars Imagery for the PDS Image Atlas

Authors
Steven Lu, Kiri Wagstaff, Emily Dunkel, Kevin Grimes, Brandon Zhao, Jesse Cai, Shoshanna B Cole, Gary Doran, Raymond Francis, Jake Lee, Lukas Mandrake
Theme
Cross-Cutting and Others
Poster ID
42

Abstract: (Not available.)

Poster 43 Thumbnail

Auto-Calibration and High-Fidelity Virtual Observations of Remote Sensing Solar Telescopes with Deep Learning

Authors
Valentina Salvatelli, Brad Neuberg, Luiz FG dos Santos, Souvik Bose, Mark Cheung, Miho Janvier, Meng Jin, Yarin Gal, Atılım Güneş Baydin
Theme
Cross-Cutting and Others
Poster ID
43

Abstract: (Not available.)

Poster 44 Thumbnail

Compressed Image Artifact Removal: Improving Instrument Data Quality After Lossy Compression

Authors
Daniel da Silva, Alex Barrie, Barbara Thompson, Ayris Narock, Michael Kirk
Theme
Cross-Cutting and Others
Poster ID
44

Abstract: Satellite instruments are collecting more data then ever before, outpacing advances in the telemetry infrastructure that enable their transmission. A common trade-off in missions is choosing to trade data quality for increased downlink volume. Two methods of doing this are decreasing temporal / spatial / spectral resolution, and lossy compression. Lossy compression is dangerous, but image quality is extensively studied in computer vision / AI.

Poster 45 Thumbnail

Best Practices in Sharing Enhanced Data Products and Machine Learning Algorithms: Learnings from NASA Frontier Development Lab

Authors
James Parr, Madhulika Guhathakurtha, Bill Diamond
Theme
Cross-Cutting and Others
Poster ID
45

Abstract: (Not available.)

Poster 46 Thumbnail

INARA: A Bayesian Deep Learning Framework for Exoplanet Atmospheric Retrieval

Authors
Frank Soboczenski, Micahel D Himes, Molly D O'Beirne, Simòné Zorzan, Atılım Güneş Baydin, Adam Cobb, Yarin Gal, Massimo Mascaro, Daniel Angerhausen, Geronimo Villanueva, Shawn D Domagal-Goldman, Giada N Arney
Theme
Cross-Cutting and Others
Poster ID
46

Abstract: (Not available.)

Poster 48 Thumbnail

Generating AI-synthetic biosensor data for future deep space missions

Authors
Brian Wang, Eleni Antoniadou, David Belo, Krittika D'Silva, Annie Martin, Brian Russell, Graham Mackintosh, Tianna Shaw, Frank Soboczenski
Theme
Cross-Cutting and Others
Poster ID
48

Abstract: (Not available.)

Poster 49 Thumbnail

Digital Transformation AI & ML Overview

Authors
Edward McLarney
Theme
Cross-Cutting and Others
Poster ID
49

Abstract: (Not available.)