National Aeronautics and Space Administration

Physics of neural network power-law scaling

The project will streamline the archiving of information and build an explorable knowledge graph of organizational communications.

The FDDCC project is guided by a set of core objectives aimed at transforming how data is accessed, managed, and utilized across the agency.

This AI use case aims to automate the cleanup, transformation, and normalization of large institutional datasets that feed into NASA's enterprise Salesforce platform.

Ongoing development and enhancement of the NASA Technology Portfolio Management System (TechPort) "AskTechPort" generative AI tool.

Enhancement and further training of the NASA Technology Taxonomy and Target Technology Destination recommendation systems, known as T-Rex and D-Rex respectively.

This project aims to leverage data available within the NASA Open Science Data Repository to train a model or set of models that can predict patterns related space health outcomes.

Probabilistic Risk Assessment (PRA) identifies minimal cut sets - smallest combination of simultaneous failures to cause a system or mission failure.

This application aims to provide the Quality Flight Equipment Division with a strong paper Work Authorization Document (WAD) data extraction capability using AI models.

SURFAS, given the above combination of inputs for a particular vehicle, is informed by expertise of mission designers to fuse relevant data together in order to create a holistic map of the scene, combining all inform...

Future astrobiological and geochemical investigations of ocean worlds (OWs) such as Europa and Enceladus will face challenges that can be addressed through science autonomy.

In recent decades, Sub-Saharan West Africa has seen rapid and ongoing land cover change fueled by population growth and subsequent agricultural expansion and intensification.

There has been much discussion on the use of Artificial Intelligence (AI) in many fields.

This AI-driven project analyzes multi-decadal solar activity datasets to identify and quantify periodic signals that correlate with planetary orbital mechanics.

Using ML RF/CNN to detect dust in MODIS images.

Current methodologies to deploy edge AI on spacecraft face a critical cost barrier due in part by reliance on traditional real-time operating systems.

Satellite-based fire detection provides critical data for fire management, fire spread modeling, air quality forecasts, and assessments of fire impacts on ecosystems and communities.

Geostationary spectrometer based foundation model (ABI-FM) and evaluation on benefits for 3D cloud and convection related downstream tasks

Text-to-Spaceship is a NASA-led initiative exploring how artificial intelligence can transform mission development.

Using Machine Learning (ML) to determine the pupil alignment of the Wide Field Instrument of the Roman Space Telescope during the spacecraft testing in thermal vacuum conditions.

This work explores the combined use of machine learning and traditional model calibration methods to develop a high resolution (1 km) soil and vegetation parameter dataset for North America and Central America.

Due to the limits in personnel capacity, warehouse inventory contract (TRAX), and the limits in the Goddard Material Management System (MMS), EEE parts kitting and auditing is labor intensive, and error prone.

BITW is a software technology designed to detect anomalous commands and trends in telemetry beyond traditional detection methods such as command parsers and telemetry limit checkers.

This project develops software to generate pre-computed, draft analysis results for text-based software requirements artifacts on NASA programs within the scope of Independent Verification & Validation (IV&V).

This project develops software to generate pre-computed, draft analysis results for static code and code implementation artifacts on NASA programs within the scope of Independent Verification & Validation (IV&V).

This project develops software to generate pre-computed, draft analysis results for text-based test artifacts on NASA programs within the scope of Independent Verification & Validation (IV&V).

Developing hourly ocean color retrievals across North America using geostationary TEMPO instrument

SAS VISION is a convolutional neural network (CNN)-based computer vision model designed to identify previously undocumented astronomical objects such as AGNs, NGCs, and stellar sources using large volumes of XMM-Newto...

Enable a rover (or rover swarm) to autonomously collect lunar regolith, haul it to a worksite, and execute construction tasks (e.g., berms, landing pad layers, trenching, or feedstock staging for sintering/printing) u...

Leveraging AI for point cloud analysis may advance accuracy and efficiency of: •Point cloud classification, enhancing capabilities for LiDAR based autonomy and navigation.

Neutron star synthetic wave form generation

Next generation cloud and aerosol parameterizations for atmospheric models

Nitrogen dioxide retrieval using hyper-spectral imagers

The Objective of this proposal is to enable the use of state-of-the-art, experimental, artificial-intelligent (AI) microchip architectures such as the Google Coral TPU (Tensor Processing Unit) on a SmallSat platform.

Planetary Trajectory Design Using Generative AI Tools

Planted Area mapping in food insecure conflict zones

PM2.5 Estimation using MERRA2 and Advanced machine learning model over US

PM2.5 Product development using AERONET data

Prediction of Spaceflight Mass Spectrometry Chemical Information with Neural Networks

Quantifying Uncertainty and Constraining Parameterizations of Clouds in Earth System Models using NASA Observations

Rangelands Water Monitoring and Forecasting System

Reconstruction of VIIRS Level1 geolocation data

Reproducing surface irradiance and penetration depth retrievals using machine learning

RST I&T Science Data Telemetry Query

RST Image Anomaly Detection

SatVision: Precursor for a foundation model developed using MODIS surface reflectance data

Science Autonomy Applications for ExoMars/MOMA

Science Keyword Link Prediction

Self-supervised learning for modeling gamma-ray variability in blazars

Software Issue Classification with LLM

SpRInT to Advance the SOA of Intelligent Space Systems

Super Resolution to enhance climate reanalysis data

Super-Resolution for Nighttime Lights

Terrain Modeling and Landmark Navigation with Radiance Fields

Terrestrial Environmental Rapid-Replicating and Assimilation Hydrometeorological (TERRAHydro) System

The evaluation of clouds in R21C data via a ML-based MODIS simulator

Identify sources of error in an Earth system model component.

Towards Learning-based Visual Perception with GAVIN: the Goddard AI Verification and INtegration Tool Suite

Towards Learning-based Visual Perception with GAVIN: the Goddard AI Verification and INtegration Tool Suite

Our technique reconstructs a 3D model of a scene based on images taken from a single viewpoint -- via estimating a depth map of a scene based on the defocus of different objects in captured images-- illuminated with l...

We have developed LunarNRM, a novel neural surface reconstruction algorithm based on Neural Radiance Fields (NeRFs) that incorporates shadow-aware and depth-aware methodologies.

The AI-CURE project will integrate advanced AI and machine learning models to automate and standardize data curation across NASA's SMD databases.

ML base lidar radiative transfer simulation for clear sky

Cloud Parameterization to improve climate model

Cache-Augmented Generation Document Search

This AI use case addresses a critical data quality challenge within NASA's enterprise Salesforce platform by enabling NASA to accurately merge extremely large institutional datasets from multiple sources into a single...

This AI use case focuses on automating the manual review and verification process for temporary "write-in" accounts submitted by external users within NASA's agency-wide enterprise Salesforce platform.

MEDOS, a flight-tested onboard decision engine

The purpose of Lunar FM is to overcome the limitations of traditional, task-specific Machine Learning (ML) models in analyzing the vast, diverse, and long-term datasets collected by the Lunar Reconnaissance Orbiter (L...

Automatically detect features of Soliton for scientists to study the ocean.

DELTA simplifies machine learning for satellite imagery.

The core Flight System (cFS) High Performance Computing Framework (HPCF) provides an environment to support a wide variety of Science work, to inlcude AI and ML.

A machine learning approach is developed to improve the bad pixel map that masks damaged or unusable pixels in the imaging spectrometers of the Orbiting Carbon Observatory-2 and -3.

A deep learning model for accurate, data-driven cloud detection in imaging spectroscopy data.

Algorithms to plan and optimize different outcomes

Methods for atmospheric retrieval of earth science data.

Methods for analyzing 3D imaging data

Machine learning based detection of methane plumes from imaging spectroscopy data.

Autonomy software to support on board AI science capabilities for the Near Earth Asteroid Scout Mission.

AQcGAN is an air quality emulator for surface O3 and NOx concentrations.

Artificial Intelligence (AI) techniques, particularly Machine Learning (ML), have undergone significant growth in heliophysics research in recent years.

Objective: Quantify the performance of Foundation Models (FMs) for weather and climate to guide GSFC scientists in effectively integrating AI into their research.

The Transiting Exoplanet Survey Satellite (TESS) is a NASA mission focused on exploring and finding exoplanets around nearby stars using the transiting method.

Radiative transfer models for satellite data assimilation and physical atmospheric retrievals need to be both fast and accurate to fulfill operational constraints.

A series of 90+ examples of how ChatGSFC or a NASA-focused LLM can be utilized to enhance project planning and controls (PP&C) analysis and streamline project management activities.

We developed a two-stage pipeline for efficient dust devil detection in Mars rover imagery.

Goal is to identify, and potentially predict when an event (e.g., anomaly, interference, etc.) has/will occur.

The development of an automated inference tool tailored to extract key physical parameters from obscured AGN (active galactic nuclei) X-ray spectra by means of more complex physical models than ever before with machin...

Observations of neutron stars provide estimates of their mass and radius-key parameters for constraining their still uncertain equation of state.

AI has helped enable me as a systems engineer to dive deep into subjects outside my expertise.

Early data-driven analyses of ozone chemistry sensitivity primarily relied on "ratio-based" indicators to partially linearize the non-linear aspects of urban ozone chemistry, which are influenced by pollution levels,...

The ASTRA team is working to develop and mature capabilities for extensibility and science autonomy.

Here we introduce an updated version of Multimodal Earth Observation Workflow for Machine Learning (MEOW-ML), an end-to-end data fusion and artificial intelligence (AI) and machine learning (ML) framework tailored for...

Use ML to reduce noise in image of satellite SO2 retrievals, and automatically identify volcanic SO2 plumes in the images.

Our objective is to fill in blocked radar data using Convolutional Neural Networks (CNN).

Implementation of models of the magnetospheric CUSP using in-situ ion flux data from ESA's Cluster mission and DMSP spacecraft.

Testing various ML algorithms to model the magnetometer offset values at points in the orbit where the traditional methods are not available.

To develop an AI/ML forecast water quality model initial targeted for the Chesapeake bay.

Neural network forward models were developed to predict subsurface vertical profile of instantaneous photosynthetically available radiation (IPAR) for both open ocean and coastal waters.

A chatgpt-like prompt query interface that uses large language models to extract intent from chat query to determine spatial, temporal and science variable filters.

Funded by the Office of the Chief Science Data Officer, the Lunar Foundation Model (LFM) is a joint effort between GSFC, MSFC, and IBM that will harness a large and diverse array of multi-modal datasets from recent mi...

ML generator for cloudy and precipitating subcolumns to be used in global climate models.

This project aims at expanding the training of two Convolutional Neural Networks (CNNs) that we have already developed to obtain a more efficient, more accurate, and least-biased CNN model for segmenting coronal holes...

Diffusion models such as DeepMind's GenCast have demonstrated powerful performance in terrestrial weather forecasting, achieving results on-par and surpassing leading medium-range numerical weather simulations.

Accurate Uncertainty Quantification (UQ) for space weather forecasts is an ever-important supplementary variable to enable accurate risk response.

Developing a CNN model to identify stellar flares in the 20-second cadence TESS data product.

The Solar Neutron TRACking (SONTRAC) instrument is designed to detect incident solar neutrons in an energy range that fills a key gap in understanding flare ion acceleration.

To develop new instrumentation that will be capable of measuring winds in planetary atmospheres.

This work focuses on creating U-Net + long short-term memory Hybrid analysis architectures for image series analysis on solar image data.

This project develops an emergent constraint emulator for future changes in water storage estimates based on the available historical record of GRACE and GRACE-FO measurements.

Development of ML based emulator for enabling what-now, what-next, and what-if analysis for flood and water quality indicators in the Chesapeake Bay, trained using Land Information System-modeled land surface variables

We propose a lightweight, computationally efficient machine learning (ML) model capable of emulating the LIS-based soil moisture and soil temperature and downscaling them from a native 10 km resolution to 1 km resolut...

A deep learning model is being trained using a subset of meteorological forcings and remote sensing observations of snow cover to reconstruct seasonal snow water content globally.

Distributed Spacecraft Autonomy (DSA) is a project developed by the National Aeronautics and Space Administration that enables distributed spacecraft systems through the development of three capabilities: scalable com...

This machine learning-driven project aims to expedite the image calibration process for Roman Wide Field Instrument (WFI) data by developing an automated calibration system.

Use an LLM to analyze existing publications and extract references to observations archived by the HEASARC

The Habitable Worlds Observatory (HWO) aims to image and thoroughly characterize exoEarths and is the highest priority of NASA as recommended by the Astro2020 Decadal Survey.

Use machine learning to increase the accuracy and range of low-order wavefront sensor.

The project's goal is to reduce the computational burden of atmospheric composition modeling at the GMAO, by building an AI emulator for the GEOS Composition Forecast model.

Neuromorphic hardware for autonomous control logic adaptation in robotics/aerospace

NASA's Mission Control currently relies on a manual CHIT (Mission Action Request) system for real-time operational decisions, a process that places a high cognitive load on flight controllers who must complete forms,...

Converts 490K+ EM32 repository files into a searchable knowledge fabric.

Allow a robotic arm(s) on the lunar surface (or inside a habitat/lander bay) to autonomously identify, retrieve, and swap between different tools and payloads, then execute a sequence of tasks (e.g., drilling, sample...

Validates engineering parts, materials, and BOMs against NASA, ASTM, and program standards.

This project is in partnership with ARC and GeoNEX under the Ecological Conservation NASA Earth Action program.

The objective of this project is to investigate the feasibility of applying deep-learning algorithms to communication-limited spacecraft, an operational domain where a slow, restricted, or intermittent downlink bottle...

Foundation Model trained on LRO WAC, NAC, and RTM imagery to assist lunar scientists with AI applications regarding the surface processes of the moon

AI to merge and analyze lunar reconnaissance data with terrain relative navigation techniques utilizing on board sensors to generate high resolution localized maps.

Various machine learning techniques (neural network, support vector machine, etc.) employed to predict bending angle of sheet metal subjected to laser forming processes and to elucidate most pertinent factors via SHAP...

NLP interface for querying legacy test reports in MAPTIS

Platform that leverages foundation models created by NASA IMPACT AI for Science to allow users to inference on the fine-tuned models and visualize the results

This work uses convolutional long short-term memory neural networks to aid in development of new tools for detecting and assessing resilient agriculural systems farm performance based on a variety of Earth and agricul...

This project leverages LLM-based tools to automate data ingestion, risk trend detection and scoring, and compliance determination across a variety of NASA applicable data.

Traditional video surveillance requires continuous human monitoring, which is resource-intensive and prone to missed events.

We are collaborating with the Magnetospheric Multiscale (MMS) mission to research Machine Learning (ML) techniques capable of predicting and detecting anomalies in spacecraft health and status data.

The Quantification of Uncertainty Analysis Toolkit (QUAnT) is a digital-twin framework that informs and guides the design process of complex, large-scale, multidisciplinary systems throughout their life cycle, while m...

We developed a data augmentation pipeline to enhance the training of vision-based navigation models for robotics, addressing the challenges of limited real-world data.

TaxiNet is a vision-based deep learning model developed to enable autonomous vehicles to follow a designated line safely during aircraft taxiing, a crucial application for assured autonomy research.

It uses Microsoft's Large Language model with scientifically curated information from NASA's VEDA (Visualization, Exploration, and Data Analysis) platform to assist users in search, discovery and analysis.

Leverage opportunities to automate overly manual analysis tasks, with a goal of amplifying productiivity and efficiency, allowing employees to free up their time for more complex tasks.

ReCAP provides an architecture for lightweight, efficient coordination of highly-capable agents in a comms-limited environment.

Onboard biosignature detection for a suite of life-detection instruments (motility, fluorescence, metabolism indicators) and summarizing the data to overcome bandwidth constraints (e.g., at Enceladus or Europa).

This work presents two complementary RAG-based chatbot systems developed for NASA's Community Coordinated Modeling Center.

This project is an AI-assisted daily workflow solution designed to streamline the scientific documentation process for physics researchers.

This project is to collect the live data from Voice Over Internet Protocol (VoIP) currently resides in NASA network.

CFI is a new pushbroom instrument with six spectral bands between the shortwave infrared (SWIR) and thermal infrared (TIR), including two channels in the mid-wave infrared (MWIR) specifically designed to detect and ch...

MADI (Modular AI for Design and Innovation) is a decentralized, open-source AI platform that identifies unexplored research "whitespace" between scientific disciplines through secure plugin architecture and interactiv...

We use Simulation Based Inference to construct 34000 artificial spectra that are representative of observed Active Galactic Nuclei X-ray spectra with NASA's NuSTAR X-ray telescope.

We plan to train an LLM on previous large code bases for high-energy astrophysics science pipeline and analysis software.

Simulates frequency of occurrence of cloud types as seen from a space-based imager from radiative fluxes.

The project develops ML/AI methods/tools to enhance the operation and sustainment of GSFC Space Network (ACCESS-managed) assets.

This project aims to develop interpretable AI/ML models to improve global snow depth retrieval from AMSR2 brightness temperature observations.

We are building a multi-agent AI copilot that lets users explore and apply the new 1-km, hourly North American Land Data Assimilation System (NLDAS) version 3 dataset through natural-language queries.

This project delivers a NASA IV&V AI Assistant powered by LibreChat with Retrieval-Augmented Generation (RAG) to provide a secure, conversational interface for NASA IV&V engineers.

This project is a collection of utility tools that generated by personnel and early adopters at the NASA IV&V Program who have explored and prototyped generative AI applications that can streamline and enhance Indepen...

XMM-GPT is a domain-specialized AI assistant built by fine-tuning Google's Open-Source FLAN family of LLMs through transfer learning on up-to-date XMM-Newton Documentation.

Global frost Martian maps derived from five remote sensing datasets and processed with tools like CNNs and other data science techniques.

Purchase card application uses ML model to suggest if a purchase may be a taggable asset or a chemical.

NTR application uses ML model to suggest a Technology Category (e.g.

Using IEEE 2874 Spatial Web standard to implement a new Hyperspace Modeling Language (HSML) and Hyperspace Transaction Protocol (HSTP) to establish communications among heterogeneous simulation platforms.

Rapid aerosol retrievals from OMPS Limb Profiler are important to monitor large wildfires and volcanic eruptions that reach the stratosphere.

Stratospheric water vapor (SWV) plays an important role in atmospheric chemistry, dynamics, and radiative forcing.

Stratospheric NO2 plays an important role in ozone photochemistry.

Processing photo images submitted to the GLOBE Program through the GLOBE Observer app.

Utilize machine learning to predict and allocate Cloud Account Allocation Plan (CAAP) cost and egress limits based on past actuals.

We are often found using optical evaluation on specimens that do not necessarily have a textbook approach to evaluation.

AI for real-time dust environment simulation and analysis

MSFC has acquired a custom dirty vacuum rated 6-axis robotic arm from Motiv space systems.