Search NASA⌕ Search

SEARCH · Search NASA

Results for “Intelligent Optimization”

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 1,009 records · Page 56
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Reinforcement learning building control approach harnessing imitation learning

Reinforcement learning (RL) has shown significant success in sequential decision making in fields like autonomous vehicles, robotics, marketing and gaming industries. This success has attracted the attention to the RL control approach for building energy systems which are becoming complicated due to the need to optimize for multiple, potentially conflicting, goals like occupant comfort, energy use and grid interactivity. However, for real world applications, RL has several drawbacks like requiring large training data and time, and unstable control behavior during the early exploration process making it infeasible for an application directly to building control tasks. To address these issues, an imitation learning approach is utilized herein where the RL agents starts with a policy transferred from accepted rule based policies and heuristic policies. This approach is successful in reducing the training time, preventing the unstable early exploration behavior and improving upon an accepted rule-based policy - all of these make RL a more practical control approach for real world applications in the domain of building controls.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

SIDDA: SInkhorn Dynamic Domain Adaptation for image classification with equivariant neural networks

Modern neural networks (NNs) often do not generalize well in the presence of a ‘covariate shift’; that is, in situations where the training and test data distributions differ, but the conditional distribution of classification labels given the data remains unchanged. In such cases, NN generalization can be reduced to a problem of learning more robust, domain-invariant features. Domain adaptation (DA) methods include a broad range of techniques aimed at achieving this; however, these methods have struggled with the need for extensive hyperparameter tuning, which then incurs significant computational costs. In this work, we introduce SInkhorn Dynamic Domain Adaptation (SIDDA), an out-of-the-box DA training algorithm built upon the Sinkhorn divergence, that can achieve effective domain alignment with minimal hyperparameter tuning and computational overhead. We demonstrate the efficacy of our method on multiple simulated and real datasets of varying complexity, including simple shapes, handwritten digits, real astronomical observations, and remote sensing data. These datasets exhibit covariate shifts due to noise, blurring, differences between telescopes, and variations in imaging wavelengths. SIDDA is compatible with a variety of NN architectures, and it works particularly well in improving classification accuracy and model calibration when paired with symmetry-aware equivariant NNs (ENNs). We find that SIDDA consistently enhances the generalization capabilities of NNs, achieving up to a ${\approx}40\%$ improvement in classification accuracy on unlabeled target data, while also providing a more modest performance gain of $\lesssim 1\%$ on labeled source data. We also study the efficacy of DA on ENNs with respect to the varying group orders of the dihedral group DN, and find that the model performance improves as the degree of equivariance increases. Finally, if SIDDA achieves proper domain alignment, it also enhances model calibration on both source and target data, with the most significant gains in the unlabeled target domain—achieving over an order of magnitude improvement in the expected calibration error and Brier score. SIDDA’s versatility across various NN models and datasets, combined with its automated approach to domain alignment, has the potential to significantly advance multi-dataset studies by enabling the development of highly generalizable models.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2019

Allocating limited deicing resources in winter snow events

As the number and severity of snowfall events continue to grow, the need to direct road maintenance intelligently during these snowfall events will also grow. In several locations, local governments lack the resources to treat all roadways completely during snow events. Furthermore, some governments utilize only traffic data to determine which roads should be treated. As a result, many schools, businesses, and government offices must be unnecessarily closed, which directly impacts the social, educational, and economic well-being of citizens and institutions. In this work, we propose a mixed integer programming (MIP) formulation to allocate resources to manage snowfall on roads using meteorological, geographical, and environmental parameters. Furthermore, given that MIPs often scale poorly with problem size, we also present a constructive heuristic designed for the problem. Finally, we evaluate the impacts of an increase in budget for winter road maintenance on snow control resources.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Contrasting Time-Frequency Representations for Unknown Waveform Detection

In real-world applications like spectrum management and interference detection, dealing with unseen electromagnetic waveforms is critical. Although some methods attempt to simulate open set data using generator models, they face challenges in generating synthetic samples for open set while simultaneously selecting an optimal discriminator for accurate classification. This results in difficulties capturing distinctive features across classes, especially in dynamic scenarios where new classes emerge. To detect unseen waveforms, we propose combining time and frequency domain features with cosine similarity loss to enhance feature distinctiveness and enabling more accurate predictions. This approach efficiently captures more comprehensive information than single-domain representations or approaches without cosine loss. Additionally, our model avoids generic feature vectors by extracting class-specific features during training, resulting in improved class representation. The experiment results show that this combined feature approach with cosine loss outperforms single-domain models and improves accuracy by 10\% over models without cosine loss.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2023

Computational optimal transport for molecular spectra: The fully continuous case

Computational optimal transport is used to analyze the difference between pairs of continuous molecular spectra. It is demonstrated that transport distances which are derived from this approach may be a more appropriate measure of the difference between two continuous spectra than more familiar measures of distance under many common circumstances. Associated with the transport distances is the transport map which provides a detailed analysis of the difference between two molecular spectra and is a key component of our study of quantitative differences between two continuous spectra. The use of optimal transport for comparing molecular spectra is developed in detail here with a set of model spectra, so that the discussion is self-contained. The difference between the transport distance and more common definitions of distance is elucidated for some well-chosen examples and it is shown where transport distances may be very useful alternatives to standard definitions of distance. The transport distance between a theoretical and experimental electronic absorption spectrum for SO 2 is studied and it is shown how the theoretical spectrum can be modified to fit the experimental spectrum better adjusting the theoretical band origin and the resolution of the theoretical spectrum. In conclusion, this analysis includes the calculation of transport maps between the theoretical and experimental spectra suggesting future applications of the methodology.

74 ATOMIC AND MOLECULAR PHYSICS↗
NASA NTRSDate not supplied

Datascope to Enable Earth Independent Medical Operations (EIMO)

BACKGROUND: NASA has amassed sixty years of knowledge and experience relevant to maintenance of crew health and performance in low earth orbit. The Apollo Program introduced the importance of ensuring progressively autonomous operational capability. Earth Independent Medical Operations (EIMO) will require a gradual shift in the balance of medical responsibility, management, and authority from terrestrial to space-based assets. Terrestrial assets will continue to be essential for pre-mission screening and planning in addition to maintenance of crew health and performance. However, new capabilities are needed to enable EIMO and the amount of data required to support these systems, and mitigate the impacts of data transmission delays and reduced bandwidth coupled with lack of cloud-like resources and on-board computing capacity that is currently unclear or operationally insufficient. OVERVIEW: The overall goal of EIMO is to develop artificial intelligence (AI)-based solutions to analyze crew health and performance data utilizing a clinical decision support system (CDSS) to provide crew medical officers (CMO) with the equivalent of real-time, on-board medical consults. The EIMO ecosystem is envisioned as a “system of systems” where embedded reference databases and real-time data streams from multiple input vectors continuously and seamlessly assess crew health and performance. EIMO will be designed to make recommendations to the CMO using multi-modal AI-based natural language processing and machine learning methods with interoperability to push/pull data within and between multiple vehicle and habitat architectures. DISCUSSION: Data flows and storage/retrieval capacity are severely constrained during space missions and the challenges will become even greater during exploration missions. Just as each past program from Mercury to the International Space Station (ISS) required rethinking the interaction between ground-based controllers and space-based crew, so too will future missions to the Moon and Mars. While the NASA High-Performance Spaceflight Computing Processor project aims to increase computational capacity by 100 times over current spaceflight computers, the projected deliverable still lags considerably behind what will be needed to enable an AI-driven CDSS. Restrictions in processing speed and data storage capacity, coupled with transmission bottlenecks and delays, necessitate definition and optimization of an integrated data architecture to enable a progressively autonomous medical capability.

Medical operations↗
DOE OSTI2023

A Machine Learning Framework to Deconstruct the Primary Drivers for Electricity Market Price Events

As the electricity grid is moving towards a 100% Renewable Energy Source Bulk Power Grid, the overall operations of the power system operations and electricity markets are changing. The electricity markets are not only dispatching resources economically but also taking into account various controllable actions like renewable curtailment, transmission congestion mitigation, and energy storage optimization to make sure the grid is operating reliably. As a result, price formations in electricity markets have become quite complex. Traditional root cause analysis and statistical approaches are rendered inapplicable to analyze and infer the main drivers behind price formation in the modern grid and markets with variable renewable energy (VRE). In this paper, we propose a machine learning analysis framework to deconstruct some primary drivers for price formation in modern electricity markets with high renewable energy and the outcomes can be utilized for various critical aspects of market design, renewable dispatch and curtailment, operations, and cyber-security applications. The framework can be applied to any ISO or market data and in this paper it is applied to open-source publicly available datasets from California Independent System Operator (CAISO) and ISO New England.

machine learning (ML), electricity markets, Renewa↗
DOE OSTI2025

Nonlinear Ensemble Filtering with Diffusion Models: Application to the Surface Quasigeostrophic Dynamics

The intersection between classical data assimilation methods and novel machine learning techniques has attracted significant interest in recent years. Here, we explore another promising solution in which diffusion models are used to formulate a robust nonlinear ensemble filter for sequential data assimilation. Unlike standard machine learning methods, the proposed ensemble score filter (EnSF) is completely training free and can efficiently generate a set of analysis ensemble members. Here, in this study, we apply the EnSF to a surface quasigeostrophic model and compare its performance against the popular local ensemble transform Kalman filter (LETKF), which makes Gaussian assumptions in the analysis step. Numerical tests demonstrate that EnSF maintains stable performance in the absence of localization and for a variety of experimental settings. We find that while LETKF maintains optimal performance in the case of linear observations of the entire state and a perfect model, EnSF shows improvements over LETKF when nonlinear observations are assimilated and the system is subject to unexpected model errors. A spectral decomposition of the analysis results in this nonlinear observation regime shows that the largest improvements over LETKF occur at large scales (small wavenumbers), where LETKF lacks sufficient ensemble spread. Overall, this initial application of EnSF to a geophysical model of intermediate complexity motivates further development of the algorithm for more realistic problems.

Artificial intelligence↗
DOE OSTI2020

Accelerating Hanford Site Cleanup through Operations Research Modeling - 20238

The Hanford Site cleanup effort will require the integration of dozens of unique facilities and processes, many of which will be first-of-a-kind in implementation and design. Each facility will be governed by its own set of operating logic, configured with a unique array of unit operations, and subject to a set of constraints that will affect its behavior. The collection of facilities have multiple points of interface, making the operations of any one facility potentially significant to the operations of other up- or downstream processes. It is therefore highly desirable to accurately predict these operations, as it allows for Site officials to identify and preempt bottlenecks and vulnerabilities before they unexpectedly inhibit the cleanup mission. With the quantity and complexity of the processes that will be on Site, building a pen-and-paper or even a spreadsheet-assisted model of the cleanup mission quickly becomes overwhelming in scope and inaccurate in execution. The Engineering organization for the Site's Tank Operations Contract (TOC) has therefore implemented the use of operations research (OR) modeling to simulate and predict future operations of Site facilities. These models are created using a discrete event simulation tool that allows for the development of detailed, versatile, and robust models. Not only can these models account for complex logical behaviors, but they can also simulate process details down to the level of vessel sizing, labor utilization, equipment reliability, and resource availability. To date, the TOC has developed OR models for several facilities on Site, including for single-shell tank (SST) farms, double-shell tank (DST) farms, the Effluent Treatment Facility (ETF), and the waste transfer system. These models have focused on identifying bottlenecks and operational constraints, and have been used to quantify the effects of implementing process changes. This latter point is particularly valuable, as it allows for several alternatives to be studied in a virtual setting before committing resources to making a change in the field. The decision to develop OR models has gained tremendous support from the Site's stakeholders and the U.S. Department of Energy (DOE) management, and has prompted the use of the tool to support additional internal and external initiatives. Recently, an initiative was proposed to use the models to help identify and provide quantitative backing for risks and opportunities for the TOC. This application of OR could not only help inform how the TOC manages its risks (e.g. quantities and types of spare parts), but could also help drive process improvements whose benefits might otherwise be hard to quantify. The models have also been used to drive the TOC's cloud computing, artificial intelligence (AI), and machine learning (ML) initiatives. These initiatives will not only improve the ability of the TOC to more rapidly respond to the needs of its customers, but it will also aid in the ability of the TOC to analyze and improve the processes it studies. Partnership with two external software development and consulting companies (Lanner and Ynformed) has furthered not only the application of AI and ML within the TOC, but has also spurred the development of new/improved software tools and platforms used by the companies. These partnerships have proven to be mutually beneficial and productive, and have set a precedent for the types of gains that can be made by exploring such options. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
NASA NTRS2002

Fiber-Optic Pressure Sensor With Dynamic Demodulation Developed

Researchers at the NASA Glenn Research Center developed in-house a method to detect pressure fluctuations using a fiber-optic sensor and dynamic signal processing. This work was in support of the Intelligent Systems Controls and Operations project under NASA's Information Technology Base Research Program. We constructed an optical pressure sensor by attaching a fiber-optic Bragg grating to a flexible membrane and then adhering the membrane to one end of a small cylinder. The other end of the cylinder was left open and exposed to pressure variations from a pulsed air jet. These pressure variations flexed the membrane, inducing a strain in the fiber-optic grating. This strain was read out optically with a dynamic spectrometer to record changes in the wavelength of light reflected from the grating. The dynamic spectrometer was built in-house to detect very small wavelength shifts induced by the pressure fluctuations. The spectrometer is an unbalanced interferometer specifically designed for maximum sensitivity to wavelength shifts. An optimum pathlength difference, which was determined empirically, resulted in a 14-percent sensitivity improvement over theoretically predicted path-length differences. This difference is suspected to be from uncertainty about the spectral power difference of the signal reflected from the Bragg grating. The figure shows the output of the dynamic spectrometer as the sensor was exposed to a nominally 2-kPa peak-to-peak square-wave pressure fluctuation. Good tracking, sensitivity, and signal-to-noise ratios are evident even though the sensor was constructed as a proof-of-concept and was not optimized in any way. Therefore the fiber-optic Bragg grating, which is normally considered a good candidate as a strain or temperature sensor, also has been shown to be a good candidate for a dynamic pressure sensor.

Lekki, John D.↗
DOE OSTI2023

A Modified Maximum Entropy Inverse Reinforcement Learning Approach for Microgrid Energy Scheduling

Increasing popularity of integrating distributed energy resources (DERs) into the power system brings a challenge to optimize the microgrid dispatch policy. The reinforcement learning methods suffer from a long-time problem with the theoretical assumption of the objective/reward function for the microgrid system. Although the traditional inverse reinforcement learning (IRL) approaches can solve this problem to some extent, they encounter a limitation of complex computations for state visitation frequency in the large and continuous state space. To alleviate this limitation, we propose a modified maximum entropy IRL (MMIRL) method to extract the reward function from the expert demonstrations for solving the microgrid energy scheduling problem. The proposed MMIRL algorithm is promising in recovering the reward function and learning the dispatch policy compared to conventional approaches. Case studies are performed in an energy arbitrage problem and a microgrid system with DERs. Results substantiate that the proposed MMIRL approach can learn the dispatch policy with more than 99% efficiency and outperforms other comparative methods.

artificial intelligence, reinforcement learning, m↗
DOE OSTI2021

Deep Reinforcement Learning Enabled Physical-Model-Free Two-Timescale Voltage Control Method for Active Distribution Systems

Active distribution networks are being challenged by frequent and rapid voltage violations due to renewable energy integration. Conventional model-based voltage control methods rely on accurate parameters of the distribution networks, which are difficult to achieve in practice. This paper proposes a novel physical-model-free two-timescale voltage control framework for active distribution systems. To achieve fast control of PV inverters, the whole network is first partitioned into several subnetworks using voltage-reactive power sensitivity. Then, the scheduling of PV inverters in the multiple sub-networks is formulated as Markov games and solved by a multi-agent soft actor-critic (MASAC) algorithm, where each subnetwork is modeled as an intelligent agent. All agents are trained in a centralized manner to learn a coordinated strategy while being executed based on only local information for fast response. For the slower time-scale control, OLTCs and switched capacitors are coordinated by a single agent-based SAC algorithm using the global information with considering control behaviors of the inverters. Particularly, the two-level agents are trained concurrently with information exchange according to the reward signal calculated from the data-driven surrogate model. Comparative tests with different benchmark methods on IEEE 33-and 123-bus systems and 342-node low voltage distribution system demonstrate that the proposed method can effectively mitigate the fast voltage violations and achieve systematical coordination of different voltage regulation assets without the knowledge of accurate system model.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2025

Self-Driving Microscopy for AI/ML-Enabled Physics Discovery and Materials Optimization

Materials are the bedrock of economy and foundation for all real-world technologies. The viability of space travel, grid energy storage, solar to fuels conversion, methane removal, and photovoltaic energy solutions hinge on the discovery and optimization of novel materials and rapid scaling toward manufacturing. The last 20 years have seen an exponential growth in the theoretical predictive capability for crystalline materials and small molecules. However, it is only in the last five years that we have seen the rapid expansion of high-throughput synthesis enabled by laboratory robotics and microfluidics, as well as a resurgence of combinatorial synthesis (Abolhasani and Kumacheva 2023; Epps and Abolhasani 2021; Jiang et al. 2022; Rajan 2008; Soldatov et al. 2021; Szymanski et al. 2023). Combinatorial synthesis, microfluidics, and ultimately dip-pen megalibraries have demonstrated the ability to “write” multicomponent nanomaterials at high throughput scale, generating millions of material examples in the 3D, 4D, and 5D composition spaces (Chen et al. 2016, 2019; Jibril et al. 2022).

36 MATERIALS SCIENCE↗
NASA NTRSDate not supplied

NASA’s Moon Trek Portal: New Capabilities Supporting Mission Planning and Engagement

Introduction: NASA’s Moon Trek (https://trek.nasa.gov/moon/) is one of a growing number of interactive, browser-based, online portals for planetary data visualization and analysis produced by NASA’s Solar System Treks Project (SSTP). Moon Trek continues to be enhanced with new data and new capabilities enabling it to facilitate the planning and conducting of upcoming lunar missions by NASA, its commercial partners, and its international partners, as well as advancing its role as a valuable outreach tool. A Comprehensive Online Web Portal: Developed at NASA’s Jet Propulsion Laboratory (JPL) and managed as a project of NASA’s Solar System Exploration Research Virtual Institute (SSERVI) at NASA Ames Research Center, Moon Trek is a browser-based web portal. The portal provides easy-to-use tools for browsing, data layering, data product blending, and feature search among thousands of data products covering topography, mineralogy, elemental abundance, geology, and much more. Visualizations are provided in var-ious map projections, interactive 3D viewing, and in virtual reality. Using an in-house stereo workflow, SSTP is able to produce new NAC-based high-resolution mosaics and DEMs. Diverse Applications for Lunar Exploration: Baseline analytic tools available to all users include dis-tance measurement, elevation profiling, sun angle calculation, and 3D print file generation. More advanced account-level tools allow users to perform more computationally intensive analyses. These include ray-traced lighting analysis for user-specified areas over user-specified time/date ranges and time intervals, electro-static surface potential analysis, subsetting of large data products, slope analysis, and Lunar Laser Ranging geometry calculation. Artificial intelligence (AI) and ma-chine learning (ML) based tools have been implemented for crater detection and hazard analysis, boulder detection and hazard analysis, and rockfall detection. New Tools Facilitating Exploration: Additional, new tools have recently been added and others are in development, offering even greater functionality in con-ducting analyses of potential landing sites and areas of surface operations. The new Line-of-Sight tool facilitates communications planning between locations on the lunar surface, between any given site on the lunar surface and a specified ground station on the Earth, and between a site on the lunar surface and a relay asset in lunar orbit, all taking into account local lunar topography. The new Data Plotter tool provides both tabular and graphical representations of pixel values along a user specified path for a growing number of data products. The new NAC Finder tool will identify and pro-vide access to NAC images that intersect a user-defined path or bounded area. The SSTP development team is looking to leverage the capabilities of its existing AI and ML crater, boulder, and rockfall detection and analysis tools, and extend that technology to a generalized feature detector that can be trained on instances of specific types of landforms and then search the lunar surface for more examples of such features. New traverse planning tools are being developed with use cases in generalized concept studies and specific mission planning in mind. These will facilitate finding optimal traverse paths based on constraints such as slope, lighting, hazard avoidance, and communications. These will be complemented by new traverse visualization capabilities. Users will be able to interactively ride along with a rover, examining 3D views of the terrain while adjusting camera height and viewing angle along with selecting different data layer overlays to drape across the terrain. Engaging the Public: The capabilities being developed for mission planning are being leveraged to further enhance Moon Trek’s proven utility as a valuable public outreach resource. This includes providing multiple lev-els of engagement with different points of entry. At its simplest level, promoting understanding through visualization, media and the public will be able to easily visualize and conduct their own exploration of lunar sites targeted by NASA and its partners. For a more in-depth experience, we are working with our stakeholders to promote understanding through interaction by extend-ing our current landing site and traverse analysis capabilities, making simplified access to these tools available to those who want to explore more deeply key factors in planning a mission through interactive and possibly even gamified experiences. The highest degree of public outreach, focusing on engagement through scientific participation, could be achieved through our work with missions and the NASA Office of the Chief Scientist on Moon Trek’s extension as a tool with specialized capabilities for facilitating citizen science. In such scenarios, participants become members of extended mis-sion science teams, using dedicated and integrated interfaces to analyze mission data to help answer questions key to lunar science and exploration. We are work-ing with NASA’s Office of Communications, museums, planetariums, and the media to help them easily integrate accurate, detailed visualizations of NASA’s lunar destinations and exploration into their content/productions and to engage diverse audiences in diverse venues.

Moon Trek↗
DOE OSTI2021

A Reconfigurable Neural Network ASIC for Detector Front-End Data Compression at the HL-LHC

Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains in the amount of data to be transported from the detector to off-detector logic where trigger decisions are made. We demonstrate that a neural network (NN) autoencoder model can be implemented in a radiation-tolerant application-specific integrated circuit (ASIC) to perform lossy data compression alleviating the data transmission problem while preserving critical information of the detector energy profile. For our application, we consider the high-granularity calorimeter from the Compact Muon Solenoid (CMS) experiment at the CERN Large Hadron Collider. The advantage of the machine learning approach is in the flexibility and configurability of the algorithm. By changing the NN weights, a unique data compression algorithm can be deployed for each sensor in different detector regions and changing detector or collider conditions. To meet area, performance, and power constraints, we perform quantization-aware training to create an optimized NN hardware implementation. The design is achieved through the use of high-level synthesis tools and the hls4ml framework and was processed through synthesis and physical layout flows based on a low-power (LP)-CMOS 65-nm technology node. The flow anticipates 200 Mrad of ionizing radiation to select gates and reports a total area of 3.6 mm 2 and consumes 95 mW of power. The simulated energy consumption per inference is 2.4 nJ. Furthermore, this is the first radiation-tolerant on-detector ASIC implementation of an NN that has been designed for particle physics applications.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Even Higher-Level Synthesis: An Exploration of AI Hardware Accelerators using HLS4ML

With the rise of artificial intelligence, the popularization of deep learning, and a constantly evolving industry, the demand for flexible and efficient tools has never been greater. As algorithms grow more complex, their runtime and energy consumption increase exponentially. Customized hardware accelerators, long used for specific mathematical operations, remain essential for managing modern applications' computational and power demands. Hardware accelerators can speed up complex computations by orders of magnitude, but their manual design and verification processes are often challenging and time-consuming. High-Level Synthesis (HLS) provides a solution by transforming high-level algorithm descriptions, typically written in C++ or SystemC, into synthesizable RTL suitable for hardware implementation. This approach reduces development time for RTL engineers while offering flexibility beyond what traditional handwritten RTL can provide. We extended this capability to the machine-learning domain with the open-source framework hls4ml, which allows neural networks trained in Python frameworks like Tensorflow or PyTorch to be synthesized into efficient hardware representations for the traditional FPGA and ASIC flows. This breakthrough addresses the growing need for reduced design turnaround and easy verification of ML hardware accelerators with low latency and power efficiency constraints. During this tutorial, we will demonstrate how Python complements HLS by simplifying the ML design process, bridging the gap between software and hardware development. Attendees will explore how we translate neural networks modeled in Python into fixed-point C++ models suitable for HLS workflows. We will dive into strategies like Value-Range Analysis and Quantization-Aware Training, which optimize these designs for deployment and evaluate their accuracy, power consumption, and energy efficiency. To exemplify these concepts, experts from Fermilab will share their experiences applying this technology to high-energy physics experiments, where real-time, low-latency processing is critical. Over the years, Fermilab engineers have demonstrated how deep neural networks, optimized for hardware using hls4ml, can meet the stringent requirements of trigger systems at the CERN Large Hadron Collider. These systems rely on rapid decision-making to process immense data volumes while retaining only the most relevant events for further analysis. The application of hls4ml has also been extended to innovative technologies like smart pixel arrays. These smart pixels integrate ML inference capabilities directly into sensor devices, enabling localized data processing at the pixel level. This approach drastically reduces the need to transmit raw data to external processing units, significantly decreasing power consumption and latency. By embedding neural networks within the pixel architecture, the smart pixels can identify and prioritize relevant data in real time, providing a highly efficient solution for edge computing in scenarios such as particle detectors and imaging systems. Fermilab's work highlights the potential of hardware-accelerated ML in scenarios where both speed and power efficiency are mission-critical. Through this tutorial, attendees will gain valuable insights into the challenges and solutions of deploying ML in hardware. Understanding how HLS and hls4ml streamline the development of neural network-based hardware accelerators is fundamental for the industry's future. Participants will learn how these technologies are shaping the future of AI and scientific computing.

Di Guglielmo, Giuseppe [Fermilab]↗
DOE OSTI2019

SDN-NGenIA, a Software Defined Next Generation Integrated Architecture for HEP and Data Intensive Science

This project aims to enable the LHC and other leading programs in high energy physics and other global science domains funded by the DOE to operate with a new level of efficiency and control, through the development of a Next Generation Integrated Architecture (NGenIA) based on intelligent software defined network (SDN)-driven network systems coupled to high throughput applications. While the initial focus is on the challenging LHC use case, the systems and products being developed are general, and apply to many fields of data intensive science ranging from astrophysical sky surveys to bioinformatics and earth observation, as well as other organizations facing the challenges of extracting knowledge from distributed multi-petabyte data stores. A central concept in this development program is a new paradigm ‘consistent network operations’ among widely distributed computing and storage facilities, where stable high throughput flows, at set rates, cross load-balanced network paths, up to flexible high water marks that are adjusted in real time to accommodate other network traffic. The large smooth flows are launched and managed by SDN services that act in concert with the experiments’ site-resident data distribution and management systems, to meet the expanding needs of the science programs. The technical goals include the construction of autonomous, intelligent site-resident services that interact dynamically with network-resident services, and with the science programs’ principal data distribution and management tools, to request or command network resources in support of high throughput petascale to exascale workflow Specific work items include: 1. Developing compact Data Transfer Nodes (DTNs) with auto-tuning functions that support data transfer rates in the 100 Gbps to the 1 Tbps range when used with high throughput data transfer applications 2. Deep site orchestration among virtualized clusters, storage subsystems and subnets to successfully coschedule CPU, storage and network resources 3. Science-program designed site architectures, operational modes, and policy and resource usage priorities, adjudicated across multiple network domains and virtual organizations, using the orchestration functions and methods 4. Seamlessly extending end-to-end operation across both extra-site and intra-site boundaries through the use of Open vSwitch (OVS), FireQoS + next generation Science DMZs 5. Novel methods of system integration that enable granular control of extreme scale long distance transfers through flow matching of scattered source-destination address pairs to multi-domain dynamic circuits 6. Funneling massive sets of streams to DTNs at the site edge hosting petascale buffer pools configured for flows of 100 Gbps and up, exploiting state of the art data transfers 7. Adaptive scheduling based on pervasive end-to-end monitoring, including DTN or compute-node resident agents providing comprehensive end-system profiling 8. Developing unsupervised and supervised machine learning and modeling methods to optimize the workflow involving terabyte to multi-petabyte datasets.

47 OTHER INSTRUMENTATION↗
DOE OSTI2026

Proliferated Resilient Economical Half-Meter Aperture Space Telescopes (PREEMPT)

The PREEMPT LDRD was motivated by a major national need for better space-based imaging systems that are both high performing and affordable for the US Government. Current optical payloads for intelligence, surveillance and reconnaissance (ISR) and space domain awareness (SDA) missions can cost hundreds of millions of dollars and take years to develop, which makes it difficult to build the large constellations required for persistent, world-wide coverage. To address this, the team aimed to advance a different kind of large-aperture (>25 cm) telescope, called a monolithic telescope, in which key optical surfaces are built into a single piece of fused silica. This design greatly reduces payload and spacecraft complexity, the need for precision focus actuators, improves mechanical and thermal robustness, and lowers cost when compared with traditional Cassegrain telescopes that rely on many precisely aligned components. The project focused on five primary thrusts. The first thrust was to advance the concept of a V10, 25 cm, monolithic telescope forward from optical design to flight-ready stage. This was accomplished in partnership with Optimax, who delivered the first test unit in the early stages of the LDRD. The team developed several technologies necessary for this optic to be integrated into a flight demonstration. These include carbon fiber housings, highly detailed structural and thermal models and stress-reducing elastic averaging Hirth groove designs. These technologies resulted in a successful maturation of the optic, which is now slated to fly in late 2026/early 2027 for a demonstration mission. The second thrust was to advance the manufacturability of these optics. In collaboration with NIF’s optical manufacturing shop, we reduced polishing time from 480 hours to 65 hours through the implementation of optimized processes and new tools. The NIF team utilized a conceptual V8 (18 cm) optic to demonstrate this optimization, though it can be applied to the rest of the monolithic optic portfolio. Third, the team developed the first conceptual V20 (50 cm) payload, which is slated to be the next generation of LLNL optical payload systems. A set of structural, dynamic and thermal simulations were performed to identify potential challenges in the future development of this payload. Early-stage simulations suggest the payload is feasible, though thermal management will be the key focus area to maintain optimal performance. Fourth, a non-linear model of Viton was developed, to further enhance the reliability of our structural and dynamic models for future payloads. Viton acts as the primary interface material between the optic and its housing. Lastly, the team focused on successfully displaying the feasibility of using additively manufactured metal composites for optical space payloads. The team successfully demonstrated layer by layer deposition of Al-SiC composites, which have highly tunable structural and coefficient of thermal expansion (CTE) properties. These are crucial for optical payloads because CTE mismatch is one of the causes for degraded optical performance for telescopes in orbit. Overall, the work showed that monolithic telescopes could become a practical, lower-cost path to high-resolution space imaging for both national security and scientific missions.

42 ENGINEERING↗