Search NASA⌕ Search

SEARCH · Search NASA

Results for “Learning algorithms”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 523 records · Page 29

Adaptive Stress Testing: Using Reinforcement Learning to Find Failures in Safety-Critical Systems

Emerging applications in artificial intelligence, such as driverless cars and autonomous aircraft promise to be more efficient, cheaper to operate, and always available. However, ensuring the safety of these systems remains a major challenge to their certification and adoption. These autonomous systems are expected to routinely make safety-critical decisions where failures can have serious consequences including loss of life and property. Testing and validation techniques aim to identify and diagnose potential failures before the system is deployed. However, finding failure scenarios in autonomous systems can be very challenging due to high-dimensional and continuous state spaces, interaction with large environments over many time steps, and the rarity of failures. This talk presents Adaptive Stress Testing (AST), a simulation-based testing framework for finding the most likely path to a failure event of a safety-critical system. The key idea of AST is that stress testing can be formulated as a Partially Observable Markov Decision Process (POMDP), which enables reinforcement learning techniques to be used for finding failure events. Reinforcement learning algorithms can efficiently explore the search space and have been shown to scale to very large systems. We present applications of AST to find failures in various safety-critical systems including the aircraft collision avoidance systems, autonomous cars, and small unmanned aerial vehicles.

autonomous vehicles↗

NASA Earth: Synthetic Spectranomics - Deep Learning of Surface 3-D Geometry, Chemistry, and Hyperspectra to Inform Next-generation Land Models

Machine and deep learning (ML/DL) have transformed our approach to Earth observation and system modeling (EOSM), unifying both in view of ML/DL models as a form of data assimilation (DA). Trained on diverse Earth observation records, detailed physical models, or hybrids of both in physics-informed machine learning, ML/DL may improve upon existing Earth system model (ESM) formulations while creating entirely new classes of models. One important application of deep learning is observation synthesis, allowing ESM developers to prepare for the increased spatial, temporal, and spectral/polar resolution of proposed future observing systems. This may involve the retrospective application of learning algorithms to existing observational records, with or without physical radiative transfer models, to co-inform mission planning and ESM development while providing a degree of data continuity for new missions.

Adam Erickson↗

California & Oregon Ecological Forecasting: Detecting and Forecasting Fog Occurrence, Frequency, and Change to Support Coast Redwood (Sequoia sempervirens) Habitat Assessments

Fog and low clouds play an important role in providing moisture to coastal ecosystems. Coast redwood (Sequoia sempervirens) forests are currently distributed along a narrow strip of coastline in California and Oregon and rely on the presence of marine fog for moisture availability during the dry season (June-October). Recent time series analyses presented an uncertain future of fog frequency; however, a decline in fog presence may have adverse effects on the coast redwood habitat. To support Save the Redwoods League, a non-profit organization dedicated to coast redwood forest management, the team analyzed hourly fog data from the Geospatial Operational Environmental Satellite 17 (GOES-17) Advanced Baseline Imager (ABI) and daily cloud cover data from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra satellite. To explore present day fog longevity, GOES-17 was utilized to map the number of fog hours per day for the 2019 and 2020 dry seasons. The MODIS cloud flag was used to map the presence or absence of daily fog, which was summarized to create a monthly fog frequency dataset and identify trends in fog presence between 2000-2020. Both datasets were used as inputs into the random forest machine learning algorithm to identify climatic drivers of fog presence and longevity over the landscape. The present-day models suggested that daily temperature difference is a driving force behind fog presence and longevity. Trends in fog presence from 2000-2020 indicated great interannual variability. Finally, fog presence was modeled under a 2080 climate projection to shed light on the future of fog presence under a projected warmer climate. Model results projected an overall decline in fog presence during the dry season in the 2080s. Decreased fog presence as a result of increased temperature difference under a warmer climate remains to be a topic of investigation as to the impact on future redwood habitat suitability.

DEVELOP Project Summary↗

California & Oregon Ecological Forecasting: Detecting and Forecasting Fog Occurrence, Frequency, and Change to Inform Coast Redwood (Sequoia sempervirens) Habitat Assessments

Fog and low clouds play an important role in providing moisture to coastal ecosystems. Coast redwood (Sequoia sempervirens) forests are currently distributed along a narrow strip of coastline in California and Oregon and rely on the presence of marine fog for moisture availability during the dry season (June-October). Recent time series analyses presented an uncertain future of fog frequency; however, a decline in fog presence may have adverse effects on the coast redwood habitat. To complement ongoing work by Save the Redwoods League, a non-profit organization dedicated to coast redwood forest management, the team analyzed hourly fog data from the Geospatial Operational Environmental Satellite 17 (GOES-17) Advanced Baseline Imager (ABI) and daily cloud cover data from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra satellite. To explore present day fog longevity, GOES-17 was utilized to map the number of fog hours per day for the 2019 and 2020 dry seasons. The MODIS cloud flag was used to map the presence or absence of daily fog, which was summarized to create a monthly fog frequency dataset and identify trends in fog presence between 2000-2020. Both datasets were used as inputs into the random forest machine learning algorithm to identify climatic drivers of fog presence and longevity over the landscape. The present-day models suggested that daily temperature difference is a driving force behind fog presence and longevity. Trends in fog presence from 2000-2020 indicated great interannual variability. Finally, fog presence was modeled under a 2080 climate projection to shed light on the future of fog presence under a projected warmer climate. Model results projected an overall decline in fog presence during the dry season in the 2080s. Decreased fog presence as a result of increased temperature difference under a warmer climate remains to be a topic of investigation as to the impact on future redwood habitat suitability.

DEVELOP Technical Paper↗

Impact of COVID-19 Restrictions on Atmospheric Concentrations of O3 and NO2 Across the Globe

We use a machine learning algorithm combining information from the NASA GEOS composition forecast (GEOS-CF) model and surface observations of nitrogen dioxide (NO2) and ozone (O3) at more than 5,000 observation sites to assess the impact of COVID-19 restrictions on surface air quality in 46 countries. Our methodology removes the compounding impacts of meteorology, seasonality and atmospheric chemistry on air pollution, thus allowing for a quantitative estimate of the change in surface air quality following COVID-19 containment measures. Compared to GEOS-CF model predictions that do not include emission reductions related to COVID-19 restrictions, surface observations show a drop in surface NO2 of up to 60% after the implementation of lockdowns. Average NO2 concentrations between February 2020 to June 2020 were 18% lower than business as usual. The earliest and strongest declines are observed over China, followed by Europe and the US. While NO2 concentrations over China recovered within 2 months, the recovery has been slower over Europe and the US. The impact of COVID-19 restrictions on O3 is complicated by non-linear atmospheric chemistry. Locally, O3 can show a short-term increase of up to 50% as a result of the decrease in NO2, which leads to a reduction in night time titration. However, this effect is offset by a decrease in photochemical production during the day. Our results indicate that these two competing processes resulted in a net zero change in average surface ozone during the first 5 months of the pandemic. The results also indicate that the reduced photochemical production becomes increasingly important over time. Our analysis is based on surface observations and model simulations available in near real-time, and we will present an up-to-date view of the short and medium-term impacts of COVID-19 restrictions on air quality around the world.

COVID-19↗

From Low-Cost Sensors to High-Quality Data: A Review of Challenges and Summary of Best Practices for Effectively Using Low-Cost Particulate Matter Mass Sensors

Low-cost sensors for particulate matter mass (PM) enable spatially dense, high temporal resolution measurements of air quality that traditional reference monitoring cannot. Low-cost PM sensors are especially beneficial in low and middle-income countries where few, if any, reference grade measurements exist and in areas where the concentration fields of air pollutants have significant spatial gradients. Unfortunately, low-cost PM sensors also come with a number of challenges that must be addressed if their data products are to be used for anything more than a qualitative characterization of air quality. The various PM sensors used in low-cost monitors are all subject to biases and calibration dependencies, corrections for which range from relatively straightforward(e.g. meteorology, age of sensor) to complex (e.g. aerosol source, composition, refractive index). The methods for correcting and calibrating these biases and dependencies that have been used in the literature likewise range from simple linear and quadratic models to complex machine learning algorithms. Here we review the needs and challenges when trying to get high-quality data from low-cost sensors. We also present a set of best practices to follow to obtain high-quality data from these low-cost sensors.

low-cost sensors↗

Exploring Transfers Between Earth-Moon Halo Orbits via Multi-Objective Reinforcement Learning

Multi-Reward Proximal Policy Optimization, a multi-objective deep reinforcement learning algorithm, is used to examine the design space of low-thrust trajectories for a SmallSat transferring between two libration point orbits in the Earth-Moon system. Using Multi-Reward Proximal Policy Optimization, multiple policies are simultaneously and efficiently trained on three distinct trajectory design scenarios. Each policy is trained to create a unique control scheme based on the trajectory design scenario and assigned reward function: a unique combination of weights scaling competing objectives that guide the spacecraft to the target mission orbit, incentivize faster flight times, and penalize propellant mass usage. Then, the policies are evaluated on the same set of perturbed initial conditions in each scenario to generate the propellant mass usages, flight times, and state discontinuities from a reference trajectory for each control scheme. This solution space of low-thrust trajectories for a SmallSat is used to examine the multi-objective trade space for the trajectory design scenario. By autonomously constructing the solution space, insights into the required propellant mass, flight time, and transfer geometry are rapidly achieved.

Christopher J Sullivan↗

Educational and Scientific Applications of Climate Model Diagnostic Analyzer

Climate Model Diagnostic Analyzer (CMDA) is a web-based information system designed for the climate modeling and model analysis community to analyze climate data from models and observations. CMDA provides tools to diagnostically analyze climate data for model validation and improvement, and to systematically manage analysis provenance for sharing results with other investigators. CMDA utilizes cloud computing resources, multi-threading computing, machine-learning algorithms, web service technologies, and provenance-supporting technologies to address technical challenges that the Earth science modeling and model analysis community faces in evaluating and diagnosing climate models. As CMDA technology and infrastructure have matured, we have developed the educational and scientific applications of CMDA. Educationally, CMDA supported the summer school of the JPL Center for Climate Sciences in 2014, 2015, and 2016. In the summer school, the students work on group research projects where CMDA provide datasets, analysis tools, and provenance support utility tools. Each student is assigned to a virtual machine with CMDA installed in Amazon Web Services. Scientifically, we have developed several science use cases of CMDA covering various topics, datasets, and analysis types. Each of the science use cases is described in terms of a scientific goal, datasets used, the analysis tools used, scientific results discovered, an analysis result such as output plots and data files, and a link to the corresponding analysis service call with all the input arguments filled.

Bao, Qihao↗

Collaborative Communications Between a Human and a Resilient Safety Support System

Successful introductory UAM integration into the NAS will be contingent on resilient safety systems that support reduced-crew flight operations. In this paper, we present a system that performs three functions: 1) monitors an operator’s physiological state; 2) assesses when the operator is experiencing anomalous states; and 3) mitigates risks by a combination of dynamic, context-based unilateral or collaborative dynamic function allocation of operational tasks. The monitoring process receives high data-rate sensor values from eye-tracking and electrocardiogram sensors. The assessment process takes these values and performs a classification that was developed using machine learning algorithms. The mitigation process invokes a collaboration protocol called DFACC to which, based on context, performs vehicle operations that the operator would otherwise routinely execute. This system has been demonstrated in a UAM flight simulator for an operator incapacitation scenario. The methods and initial results as well as relevant UAM and AAM scenarios will be described.

Advanced Air Mobility,↗

Global Impact of COVID-19 Restrictions on the Atmospheric Concentrations of Nitrogen Dioxide and Ozone

Social-distancing to combat the COVID-19 pandemic has led to widespread reductions in air pollutant emissions. Quantifying these changes requires a business-as-usual counterfactual that accounts for the synoptic and seasonal variability of air pollutants. We use a machine learning algorithm driven by information from the NASA GEOS-CF model to assess changes in nitrogen dioxide (NO2) and ozone (O3) at 5,756 observation sites in countries from January through June 2020. Reductions in NO2 coincide with timing and intensity of COVID-19 restrictions, ranging from 60% in severely affected cities (e.g., Wuhan, Milan) to little change (e.g., Rio de Janeiro, Taipei). On average, NO2 concentrations were (13-23) % lower than business as usual from February 2020 onward. China experienced the earliest and steepest decline, but concentrations since April have mostly recovered and remained within 5% to the business-as-usual estimate. NO2 reductions in Europe and the US have been more gradual with a halting recovery starting in late March. We estimate that the global NOx (NO+NO2) emission reduction during the first 6 months of 2020 amounted to 3.1 (2.6-3.6) TgN, equivalent to 5.5(4.7-6.4) % of the annual anthropogenic total. The response of surface O3 is complicated by competing influences of non-linear atmospheric chemistry. While surface O3 increased by up to 50% in some locations, we find the overall net impact on daily average O3 between February -June 2020 to be small. However, our analysis indicates a flattening of the O3 diurnal cycle with an increase in nighttime ozone due to reduced titration and a decrease in daytime ozone, reflecting a reduction in photochemical production.

Christoph A Keller↗

QuantifyML: How Good is My Machine Learning Model?

We present QuantifyML, which applies model counting to assess the learn ability, safety, and robustness of machine learning models. Typically, the efficacy of machine learning models is determined by computing their accuracy statistically on test datasets. However, this may be misleading, if the test data is not representative of the problem that is being studied. With QuantifyML we aim to precisely quantify the extent to which machine learning models have learned and generalized from the given data. In QuantifyML, a trained model is translated into aC program, which is fed to the CBMC model checking tool to produce a formula in Conjunctive Normal Form (CNF), which in turn is analyzed with state-of-the-art model counters to efficiently obtain precise countsw.r.t different outputs. QuantifyML enables i) evaluating the learnability of models by comparing the counts for the outputs to ground truth, ex-pressed as logical predicates (if available), ii) comparing the performance of different models that may be built with different machine learning algorithms (e.g., decision-trees vs. neural networks), and iii) quantifying the safety and robustness of trained models.

Deep Neural Networks↗

MULTI-OBJECTIVE REINFORCEMENT LEARNING FOR LOW-THRUST TRANSFER DESIGN BETWEEN LIBRATION POINT ORBITS

Multi-Reward Proximal Policy Optimization (MRPPO) is a multi-objective reinforcement learning algorithm used to construct low-thrust transfers between periodic orbits in multi-body systems. Previous implementations of MRPPO have relied on a predefined reference transfer to successfully train each policy. In this paper, an algorithmic modification labeled the ‘moving reference’, is introduced to autonomously construct these reference trajectories during training. With this modification, MRPPO is used to recover various low-thrust transfers between two periodic orbits in the Earth-Moon circular restricted three-body problem to solve a multi-objective optimization problem. These results are then compared with the solutions recovered via a traditional optimization formulation.

algorithm↗

Collaborative Communications Between A Human and A Resilient Safety Support System

Successful introductory UAM integration into the NAS will be contingent on resilient safety systems that support reduced-crew flight operations. In this paper, we present a system that performs three functions: 1) monitors an operator’s physiological state; 2) assesses when the operator is experiencing anomalous states; and 3) mitigates risks by a combination of dynamic, context-based unilateral or collaborative dynamic function allocation of operational tasks. The monitoring process receives high data-rate sensor values from eye-tracking and electrocardiogram sensors. The assessment process takes these values and performs a classification that was developed using machine learning algorithms. The mitigation process invokes a collaboration protocol called DFACCto which, based on context, performs vehicle operations that the operator would otherwise routinely execute. This system has been demonstrated in a UAM flight simulator for an operator incapacitation scenario. The methods and initial results as well as relevant UAM and AAM scenarios will be described.

Advanced air mobility↗

MULTI-OBJECTIVE REINFORCEMENT LEARNING FOR LOW-THRUST TRANSFER DESIGN BETWEEN LIBRATION POINT ORBITS

Multi-Reward Proximal Policy Optimization (MRPPO) is a multi-objective reinforcement learning algorithm used to construct low-thrust transfers between periodic orbits in multi-body systems. Previous implementations of MRPPO have relied on a predefined reference transfer to successfully train each policy. In this paper, an algorithmic modification labeled the ‘moving reference’, is introduced to autonomously construct these reference trajectories during training. With this modification, MRPPO is used to recover various low-thrust transfers between two periodic orbits in the Earth-Moon circular restricted three-body problem to solve a multi-objective optimization problem. These results are then compared with the solutions recovered via a gradient descent optimization scheme to validate the performance of MRPPO with the moving reference modification.

Christopher J. Sullivan↗

MULTI-OBJECTIVE REINFORCEMENT LEARNING FOR LOW-THRUST TRANSFER DESIGN BETWEEN LIBRATION POINT ORBITS

Multi-Reward Proximal Policy Optimization (MRPPO) is a multi-objective reinforcement learning algorithm used to construct low-thrust transfers between periodic orbits in multi-body systems. Previous implementations of MRPPO have relied on a predefined reference transfer to successfully train each policy. In this paper, an algorithmic modification labeled the ‘moving reference’, is introduced to autonomously construct these reference trajectories during training. With this modification, MRPPO is used to recover various low-thrust transfers between two periodic orbits in the Earth-Moon circular restricted three-body problem to solve a multi-objective optimization problem. These results are then compared with the solutions recovered via a gradient descent optimization scheme to validate the performance of MRPPO with the moving reference modification

Christopher John Sullivan↗

Exploring Transfers Between Earth-Moon Halo Orbits via Multi-Objective Reinforcement Learning

Multi-Reward Proximal Policy Optimization, a multi-objective deep reinforcement learning algorithm, is used to examine the design space of low-thrust trajectories for a SmallSat transferring between two libration point orbits in the Earth- Moon system. Using Multi-Reward Proximal Policy Optimiza- tion, multiple policies are simultaneously and efficiently trained on three distinct trajectory design scenarios. Each policy is trained to create a unique control scheme based on the trajectory design scenario and assigned reward function: a unique combination of weights scaling competing objectives that guide the spacecraft to the target mission orbit, incentivize faster flight times, and penalize propellant mass usage. Then, the policies are evaluated on the same set of perturbed initial conditions in each scenario to generate the propellant mass usages, flight times, and state discontinuities from a reference trajectory for each control scheme. This solution space of low-thrust trajectories for a SmallSat is used to examine the multi-objective trade space for the trajectory design scenario. By autonomously constructing the solution space, insights into the required propellant mass, flight time, and transfer geometry are rapidly achieved.

Mashiku, Alinda K.↗

Spacecraft Material Characterization Using Reflectance Spectra Extracted from RGB/IR Color Images

The space environment is detrimental to the exterior of the spacecraft and imposes stringent requirements on spacecraft materials. Space materials are often expected to perform for 15-20 years in the harsh space environment without significant degradation of their optical, electrical, mechanical and thermal properties. Further, as commercial activities increase in the space domain, development of predictive capability for material evolution is critical for agile adaption of novel materials. Therefore, it is very important to understand the effects of the space environments on the spacecraft materials. The Materials on the International Space Station Experiment Flight Facility (MISSE-FF) has flown numerous material samples to investigate the effect of LEO space weather exposure on the performance and durability of materials and devices. We are manifest on the MISSE-16 mission, scheduled to launch in June 2022, to fly fifteen novel and well-characterized materials in the LEO environment for a duration of six months. Changes in spectral reflectivity will be measured in real time throughout the mission with an RGB/IR camera. These time-resolved data will function as a "space-truth" reference for our team's ongoing laboratory-based space weather-material interaction experiments. Correlation of the MISSE-16 data with extensive ground testing of duplicate samples under simulated space weather conditions will enable development of fundamental chemical models for material degradation. This paper discusses preliminary results from the ground test campaign to collect the RGB/IR images for pristine and damaged materials and the development of machine learning algorithms to extract reflectance spectra from the color images.

Jainisha R. Shah↗

Detecting Satellite Laser Ranging Station Data and Operational Anomalies with Machine Learning Isolation Forests at NASA's CDDIS

The International Laser Ranging Service (ILRS) is currently composed of 45 active satellite laser ranging (SLR) stations with several more set to join the network over the next several years. Station changes and histories are logged to files, but not always in real time. Sometimes these details are not added until long after changes have been made to the station –on occasion, years later. This in addition to unexpected hardware errors and other system issues that are not immediately detected impact the products generated by analysts. The ILRS Central Bureau (CB) and NASA’s Crustal Dynamics Data Information System (CDDIS) have worked to provide tools for station engineers to use. This includes the creation of station plots which contain temperature and pressure information along with LAser GEOdynamic Satellite (LAGEOS) and LAser RElativity Satellite (LARES) tracking information that enable the monitoring of station performance and todetermine whether the station has undergone any changes. As next steps, the CDDIS is working to enhance these station performance monitoring tools through machine learning. Isolation forest is an unsupervised machine learning algorithm commonly applied to anomaly detection. In this poster, the CDDIS details the steps taken to track anomalies within SLR station performance using isolation forest with LAGEOS and LARES satellite data.

Benjamin P Michael↗