Search NASA⌕ Search

SEARCH · Search NASA

Results for “operator inference”

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 73 records · Page 4

Accelerating Hamiltonian Monte Carlo for Bayesian inference in neural networks and neural operators

Hamiltonian Monte Carlo (HMC) is a powerful and accurate method to sample from the posterior distribution in Bayesian inference. However, HMC techniques are computationally demanding for Bayesian neural networks due to the high dimensionality of the network’s parameter space and the non-convexity of their posterior distributions. Therefore, various approximation techniques, such as variational inference (VI) or stochastic gradient MCMC, are often employed to infer the posterior distribution of the network parameters. Such approximations introduce inaccuracies in the inferred distributions, resulting in unreliable uncertainty estimates. In this work, we propose a hybrid approach that combines inexpensive VI and accurate HMC methods to efficiently and accurately quantify uncertainties in neural networks and neural operators. The proposed approach leverages an initial VI training on the full network. We examine the influence of individual parameters on the prediction uncertainty, which shows that a large proportion of the parameters do not contribute substantially to uncertainty in the network predictions. This information is then used to significantly reduce the dimension of the parameter space, and HMC is performed only for the subset of network parameters that strongly influence prediction uncertainties. This yields a framework for accelerating the full batch HMC for posterior inference in neural networks. We demonstrate the efficiency and accuracy of the proposed framework on deep neural networks and operator networks, showing that inference can be performed for large networks with tens to hundreds of thousands of parameters. Finally, we show that this method can effectively learn surrogates for complex physical systems by modeling the operator that maps from upstream conditions to wall-pressure data on a cone in hypersonic flow.

Bayesian inference↗

On/Off State Classification of a Reactor Facility Using Gas Effluence Measurements

Inferring on/off operational state of a reactor facility using measurements from an independent monitoring system is critical to the assessment of its compliance to agreements. In this paper, we consider the problem of inferring on/off state of a reactor facility using measurements of Ar-41, Cs-138, and Xe-138 gas effluence types collected at the facility’s off-gas stack. We present classifiers based on thresholding measurements of individual effluence types, and then present fusers that combine their outputs or measurements. We present five fusers based on simple majority rule, Chow’s pattern recognition function, Fisher’s combined p-value statistic, physics-based Poisson radiation counts model, and correlation coefficient method. In addition, we also test five machine learning methods based on non-linear classifiers, which are available as R packages. Finally, our results show that: (i) these gas effluence measurements are effective in inferring on/off state of a reactor facility; for example, best fusers achieve ~97% detection at ~1% false alarm rate, and (ii) fusers that combine all effluence types based on physics-based models, correlation coefficient and Fisher’s method outperform simple majority rule, Chow’s fusers and machine learning methods, as well as when they are applied to individual and pairs of effluence types.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Evaluating the Interplay between Trajectory Segmentation and Mode Inference Error

Travel behavior changes are essential to transportation decarbonization. Travel diaries, consisting of sequences of trips between places, are typically used to instrument human travel behavior. However, these diaries are only as accurate as the underlying methods used to construct them. Travel diary algorithms have been a popular research topic since the advent of Global Positioning System tracking surveys. These algorithms have typically been validated using prompted recall of presegmented trips, thus disregarding the continuity of mode inference. Phone operating systems have adopted battery-conserving techniques, but the resulting data collection errors have not been studied extensively. We introduce a framework to evaluate the accuracy of trip length computations and mode inference by analyzing continuous mode-segmented trajectories for groups of trips. We then use the framework to identify the input data quality and the impact of postprocessing. Our primary inputs to this evaluation are MobilityNet, a public dataset containing information from three artificial timelines covering 15 different travel modes, and sample open-source travel diary creation algorithms from the OpenPATH project. Our framework concretely shows that the variance of the distance error drops from (0.217, 0.0848) to (o.011, 0.0407) (Android, iOS) after postprocessing. Similarly, the weighted F-scores for mode inference increase from (0.25, 0.29) to (0.60, 0.74) (iOS, Android) between random forest and geographic information system-based models. We hope that this standardized method will be adapted to evaluate other, potentially proprietary, travel diary algorithms. Finally, the results can be used to understand and improve the state of the art in the travel diary creation field.

33 ADVANCED PROPULSION SYSTEMS↗

Human-Robot Interaction Directed Research Project

Human-robot interaction (HRI) is about understanding and shaping the interactions between humans and robots (Goodrich & Schultz, 2007). It is important to evaluate how the design of interfaces and command modalities affect the human s ability to perform tasks accurately, efficiently, and effectively (Crandall, Goodrich, Olsen Jr., & Nielsen, 2005) It is also critical to evaluate the effects of human-robot interfaces and command modalities on operator mental workload (Sheridan, 1992) and situation awareness (Endsley, Bolt , & Jones, 2003). By understanding the effects of interface design on human performance, workload, and situation awareness, interfaces can be developed that support the human in performing tasks with minimal errors and with appropriate interaction time and effort. Thus, the results of research on human-robot interfaces have direct implications for design. Because the factors associated with interfaces and command modalities in HRI are too numerous to address in 3 years of research, the proposed research concentrates on three manageable areas applicable to National Aeronautics and Space Administration (NASA) robot systems. These topic areas emerged from the Fiscal Year (FY) 2011 work that included extensive literature reviews and observations of NASA systems. The three topic areas are: 1) video overlays, 2) camera views, and 3) command modalities. Each area is described in detail below, along with relevance to existing NASA human-robot systems. In addition to studies in these three topic areas, a workshop is proposed for FY12. The workshop will bring together experts in human-robot interaction and robotics to discuss the state of the practice as applicable to research in space robotics. Studies proposed in the area of video overlays consider two factors in the implementation of augmented reality (AR) for operator displays during teleoperation. The first of these factors is the type of navigational guidance provided by AR symbology. In the proposed studies, participants performance during teleoperation of a robot arm will be compared when they are provided with command-guidance symbology (that is, directing the operator what commands to make) or situation-guidance symbology (that is, providing natural cues so that the operator can infer what commands to make). The second factor for AR symbology is the effects of overlays that are either superimposed or integrated into the external view of the world. A study is proposed in which the effects of superimposed and integrated overlays on operator task performance during teleoperated driving tasks are compared

Rochlis, Jennifer↗

An Ada inference engine for expert systems

The purpose is to investigate the feasibility of using Ada for rule-based expert systems with real-time performance requirements. This includes exploring the Ada features which give improved performance to expert systems as well as optimizing the tradeoffs or workarounds that the use of Ada may require. A prototype inference engine was built using Ada, and rule firing rates in excess of 500 per second were demonstrated on a single MC68000 processor. The knowledge base uses a directed acyclic graph to represent production lines. The graph allows the use of AND, OR, and NOT logical operators. The inference engine uses a combination of both forward and backward chaining in order to reach goals as quickly as possible. Future efforts will include additional investigation of multiprocessing to improve performance and creating a user interface allowing rule input in an Ada-like syntax. Investigation of multitasking and alternate knowledge base representations will help to analyze some of the performance issues as they relate to larger problems.

Lavallee, David B.↗

Reservoir Assessment Tool 2.0: Stakeholder driven improvements to satellite remote sensing based reservoir monitoring

In light of the rapidly increasing regulation of rivers due to planned and constructed reservoirs, monitoring reservoir operations has become very crucial. The Reservoir Assessment Tool (RAT) framework was developed to monitor reservoir operations globally, using hydrological modeling and satellite observations. With feedback from stakeholders, improvements in the RAT framework are demonstrated in this study using the Mekong River Basin as an example. A novel multi-sensor reservoir area mapping technique was developed using complementary strengths of optical and SAR sensors at a 1-5 day temporal resolution, allowing the quantification of sub-weekly reservoir operations. Additionally, the skill of radar altimeters in the RAT framework was tested using the Jason-3 altimeter. Using in-situ data from three Thai reservoirs in the Mekong Basin, consistent improvements were observed as compared to the original RAT framework. A functionality for visualizing forecasted outflow was also added using historically inferred reservoir operations or stakeholder-driven target storage expectations.

Reservoirs↗

Facility Activity Inference Using Radiation Networks

We consider the problem of inferring the operational status of a reactor facility using measurements from a radiation sensor network deployed around the facility’s ventilation off-gas stack. The intensity of stack emissions decays with distance, and the sensor counts or measurements are inherently random with parameters determined by the intensity at the sensor’s location. We utilize the measurements to estimate the intensity at the stack, and use it in a one-sided Sequential Probability Ratio Test (SPRT) to infer on/off status of the reactor. We demonstrate the superior performance of this method over conventional majority fusers and individual sensors using (i) test measurements from a network of 21 NaI detectors, and (ii) effluence measurements collected at the stack of a reactor facility. We also analytically establish the superior detection performance of the network over individual sensors with fixed and adaptive thresholds by utilizing the Poisson distribution of the counts. We quantify the performance improvements of the network detection over individual sensors using the packing number of the intensity space.

Rao, Nageswara S.↗

EdgeAI: Machine learning via direct attached accelerator for streaming data processing at high shot rate x-ray free-electron lasers

We present a case for low batch-size inference with the potential for adaptive training of a lean encoder model. We do so in the context of a paradigmatic example of machine learning as applied in data acquisition at high data velocity scientific user facilities such as the Linac Coherent Light Source-II x-ray Free-Electron Laser. We discuss how a low-latency inference model operating at the data acquisition edge can capitalize on the naturally stochastic nature of such sources. We simulate the method of attosecond angular streaking to produce representative results whereby simulated input data reproduce high-resolution ground truth probability distributions. By minimizing the mean-squared error between the decoded output of the latent representation and the ground truth distributions, we ensure that the encoding layers and resulting latent representation maintains full fidelity for any downstream task, be it classification or regression. We present throughput results for data-parallel inference of various batch sizes, some with throughput exceeding 100 k images per second. We also show in situ training below 10 s per epoch for the full encoder–decoder model as would be relevant for streaming and adaptive real-time data production at our nation’s scientific light sources.

97 MATHEMATICS AND COMPUTING↗

Learning turbulent flows with generative models for super resolution and sparse flow reconstruction

Neural operators are promising surrogates for dynamical systems but when trained with standard L 2 losses they tend to oversmooth fine-scale turbulent structures. Here, we show that combining operator learning with generative modeling overcomes this limitation. We consider three practical turbulent-flow challenges where conventional neural operators fail: spatio-temporal super-resolution, forecasting, and sparse flow reconstruction. For Schlieren jet super-resolution, an adversarially trained neural operator (adv-NO) reduces the energy-spectrum error by 15 × while preserving sharp gradients at neural operator-like inference cost. For 3D homogeneous isotropic turbulence, adv-NO trained on only 160 timesteps from a single trajectory forecasts accurately for five eddy-turnover times and offers 114 × wall-clock speed-up at inference than the baseline diffusion-based forecasters, enabling near-real-time rollouts. For reconstructing cylinder wake flows from highly sparse Particle Tracking Velocimetry-like inputs, a conditional generative model infers full 3D velocity and pressure fields with correct phase alignment and statistics. These advances enable accurate reconstruction and forecasting at low compute cost, bringing near-real-time analysis and control within reach in experimental and computational fluid mechanics.

Fluid dynamics↗

VpROM: a novel variational autoencoder-boosted reduced order model for the treatment of parametric dependencies in nonlinear systems

Reduced Order Models (ROMs) are of considerable importance in many areas of engineering in which computational time presents difficulties. Established approaches employ projection-based reduction, such as Proper Orthogonal Decomposition. The limitation of the linear nature of such operators is typically tackled via a library of local reduction subspaces, which requires the assembly of numerous local ROMs to address parametric dependencies. Our work attempts to define a more generalisable mapping between parametric inputs and reduced bases for the purpose of generative modeling. We propose the use of Variational Autoencoders (VAEs) in place of the typically utilised clustering or interpolation operations, for inferring the fundamental vectors, termed as modes, which approximate the manifold of the model response for any and each parametric input state. The derived ROM still relies on projection bases, built on the basis of full-order model simulations, thus retaining the imprinted physical connotation. However, it additionally exploits a matrix of coefficients that relates each local sample response and dynamics to the global phenomena across the parametric input domain. The VAE scheme is utilised for approximating these coefficients for any input state. This coupling leads to a high-precision low-order representation, which is particularly suited for problems where model dependencies or excitation traits cause the dynamic behavior to span multiple response regimes. Moreover, the probabilistic treatment of the VAE representation allows for uncertainty quantification on the reduction bases, which may then be propagated to the ROM response. The performance of the proposed approach is validated on an open-source simulation benchmark featuring hysteresis and multi-parametric dependencies, and on a large-scale wind turbine tower characterised by nonlinear material behavior and model uncertainty.

Conditional VAEs↗

DriveSense: A Noise-Resilient Framework for Driving Mode Identification

Accurate drive mode classification is essential for enhancing the reliability and predictive maintenance of heavy-duty electric trucks. This study proposes a novel fuzzy logic-based framework, DriveSense, for real-time drive mode classification, addressing key challenges such as sensor noise, transitional behaviors, and computational efficiency. The proposed approach integrates a two-stage filtering pipeline, combining adaptive outlier removal and a dynamic Kalman filter to enhance data quality. A fuzzy inference system with smoothened trapezoidal membership functions is then applied to classify driving modes into standstill, constant speed, acceleration, and deceleration while mitigating the effects of noise and edge cases. Performance evaluation using real-world and simulated drive cycles demonstrates significant improvements in classification accuracy (up to 97.8%), F1-score (up to 0.97), and robustness against noise, while reducing false positives. Comparative analysis against baseline models, demonstrates DriveSense’s superior accuracy and generalizability across diverse driving patterns. The framework’s lightweight and interpretable fuzzy inference engine operates with low computational latency, ensuring compatibility with real-time embedded systems typical of heavy-duty electric trucks. Moreover, DriveSense models transitional behaviors through overlapping fuzzy sets and adaptive borderline classification logic, enabling smooth identification of subtle shifts such as rolling stops or gradual deceleration. These results highlight DriveSense’s potential to enhance predictive maintenance strategies, reduce downtime, and support scalable, fleet-wide diagnostics.

Kumar, Praveen [Oak Ridge National Laboratory (ORN↗

Annotated Bibliography for the MATADOR Project

The MATADOR project is focused on developing methods to infer the operational mode of facilities that have the potential to be used in weapons development programs. Our central hypothesis is that by persistent, non-intrusive monitoring of such facilities, differences between various use scenarios can be reliably discovered. The impact of success in this area is that new tools and techniques for monitoring and treaty verification would make it easier to reliably discover and document weapons development activities. This document captures the literature that will serve as a basis to approach this task. The relevant literature is divided into topical areas that relate to the various aspects of expected MATADOR project development. We have found that very little work that is directly applicable for our purposes has been published, which has motivated the development of novel methods under the project. Therefore, the manuscripts referenced in this document were selected based on their potential use as foundational blocks for the methods we anticipate developing, or so that we can understand the limitations of existing methods.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Mobile Building Energy Audit and Modeling Tools: Cooperative Research and Development (Final Report)

Broadly accessible, low cost, accurate, and easy-to-use energy auditing tools remain out of reach for managers of the aging U.S. building population (over 80% of U.S. commercial buildings are more than 10 years old*). concept3D and NREL's commercial buildings group will work to translate and extend NREL's existing spreadsheet-based energy auditing tool for a browser-friendly and mobile-computing platform. NREL will also work with concept3D to further develop a prototype geometry capture and materials inference tool operable on a smart phone/pad platform. These tools will be developed to interoperate with NREL's Building Component Library and OpenStudio energy modeling platforms, and will be marketed by concept3D to commercial developers, academic institutions and governmental agencies. concept3D is NREL's lead developer and subcontractor of the Building Component Library.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Environmental modeling and recognition for an autonomous land vehicle

An architecture for object modeling and recognition for an autonomous land vehicle is presented. Examples of objects of interest include terrain features, fields, roads, horizon features, trees, etc. The architecture is organized around a set of data bases for generic object models and perceptual structures, temporary memory for the instantiation of object and relational hypotheses, and a long term memory for storing stable hypotheses that are affixed to the terrain representation. Multiple inference processes operate over these databases. Researchers describe these particular components: the perceptual structure database, the grouping processes that operate over this, schemas, and the long term terrain database. A processing example that matches predictions from the long term terrain model to imagery, extracts significant perceptual structures for consideration as potential landmarks, and extracts a relational structure to update the long term terrain database is given.

Lawton, D. T.↗

Fuzzy logic in control systems: Fuzzy logic controller. I, II

Recent advances in the theory and applications of fuzzy-logic controllers (FLCs) are examined in an analytical review. The fundamental principles of fuzzy sets and fuzzy logic are recalled; the basic FLC components (fuzzification and defuzzification interfaces, knowledge base, and decision-making logic) are described; and the advantages of FLCs for incorporating expert knowledge into a control system are indicated. Particular attention is given to fuzzy implication functions, the interpretation of sentence connectives (and, also), compositional operators, and inference mechanisms. Applications discussed include the FLC-guided automobile developed by Sugeno and Nishida (1985), FLC hardware systems, FLCs for subway trains and ship-loading cranes, fuzzy-logic chips, and fuzzy computers.

Lee, Chuen Chien↗

Using Natural Language to Enhance Mission Effectiveness

The availability of highly capable, yet relatively cheap, unmanned aerial vehicles (UAVs) is opening up new areas of use for hobbyists and for professional-related activities. The driving function of this research is allowing a non-UAV pilot, an operator, to define and manage a mission. This paper describes the preliminary usability measures of an interface that allows an operator to define the mission using speech to make inputs. An experiment was conducted to begin to enumerate the efficacy and user acceptance of using voice commands to define a multi-UAV mission and to provide high-level vehicle control commands such as "takeoff." The primary independent variable was input type - voice or mouse. The primary dependent variables consisted of the correctness of the mission parameter inputs and the time needed to make all inputs. Other dependent variables included NASA-TLX workload ratings and subjective ratings on a final questionnaire. The experiment required each subject to fill in an online form that contained comparable required information that would be needed for a package dispatcher to deliver packages. For each run, subjects typed in a simple numeric code for the package code. They then defined the initial starting position, the delivery location, and the return location using either pull-down menus or voice input. Voice input was accomplished using CMU Sphinx4-5prealpha for speech recognition. They then inputted the length of the package. These were the option fields. The subject had the system "Calculate Trajectory" and then "Takeoff" once the trajectory was calculated. Later, the subject used "Land" to finish the run. After the voice and mouse input blocked runs, subjects completed a NASA-TLX. At the conclusion of all runs, subjects completed a questionnaire asking them about their experience in inputting the mission parameters, and starting and stopping the mission using mouse and voice input. In general, the usability of voice commands is acceptable. With a relatively well-defined and simple vocabulary, the operator can input the vast majority of the mission parameters using simple, intuitive voice commands. However, voice input may be more applicable to initial mission specification rather than for critical commands such as the need to land immediately due to time and feedback constraints. It would also be convenient to retrieve relevant mission information using voice input. Therefore, further on-going research is looking at using intent from operator utterances to provide the relevant mission information to the operator. The information displayed will be inferred from the operator's utterances just before key phrases are spoken. Linguistic analysis of the context of verbal communication provides insight into the intended meaning of commonly heard phrases such as "What's it doing now?" Analyzing the semantic sphere surrounding these common phrases enables us to predict the operator's intent and supply the operator's desired information to the interface. This paper also describes preliminary investigations into the generation of the semantic space of UAV operation and the success at providing information to the interface based on the operator's utterances.

Trujillo, Anna C.↗

The Intercalibration of Geostationary Visible Imagers Using Operational Hyperspectral SCIAMACHY Radiances

Spectral band differences between sensors can complicate the process of intercalibration of a visible sensor against a reference sensor. This can be best addressed by using a hyperspectral reference sensor whenever possible because they can be used to accurately mitigate the band differences. This paper demonstrates the feasibility of using operational Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY) large-footprint hyperspectral radiances to calibrate geostationary Earth-observing (GEO) sensors. Near simultaneous nadir overpass measurements were used to compare the temporal calibration of SCIAMACHY with Aqua Moderate Resolution Imaging Spectroradiometer band radiances, which were found to be consistent to within 0.44% over seven years. An operational SCIAMACHY/GEO ray-matching technique was presented, along with enhancements to improve radiance pair sampling. These enhancements did not bias the underlying intercalibration and provided enough sampling to allow up to monthly monitoring of the GEO sensor degradation. The results of the SCIAMACHY/GEO intercalibration were compared with other operational four-year Meteosat-9 0.65-μm calibration coefficients and were found to be within 1% of the gain, and more importantly, it had one of the lowest temporal standard errors of all the methods. This is more than likely that the GEO spectral response function could be directly applied to the SCIAMACHY radiances, whereas the other operational methods inferred a spectral correction factor. This method allows the validation of the spectral corrections required by other methods.

Doelling, David R.↗

How to select combination operators for fuzzy expert systems using CRI

A method to select combination operators for fuzzy expert systems using the Compositional Rule of Inference (CRI) is proposed. First, fuzzy inference processes based on CRI are classified into three categories in terms of their inference results: the Expansion Type Inference, the Reduction Type Inference, and Other Type Inferences. Further, implication operators under Sup-T composition are classified as the Expansion Type Operator, the Reduction Type Operator, and the Other Type Operators. Finally, the combination of rules or their consequences is investigated for inference processes based on CRI.

Turksen, I. B.↗