Search NASA⌕ Search

SEARCH · Search NASA

Results for “Adaptive tuning”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Large language models for transportation research: Methodologies, state of the art, and future opportunities

The rapid rise of large language models (LLMs) is transforming transportation research, with significant advancements emerging between 2023 and 2025, a period marked by the inception and swift growth of adopting and adapting LLMs for various transportation applications. Despite these significant advancements, however, a systematic review and synthesis of the existing literature remains lacking. This paper aims to fill this gap by providing a comprehensive review of the methodologies and applications of LLMs in transportation. We explore key applications, including autonomous driving, travel behavior prediction, and general transportation-related queries, alongside LLM methodologies such as zero- or few-shot learning, prompt engineering, and fine-tuning. From the review, critical research gaps are identified. From the methodological perspective, many of the research limitations can be addressed by integrating LLMs with existing tools and refining LLM architectures. From the application perspective, research opportunities for LLMs to address various transportation challenges are also explored. By synthesizing these findings, this review not only presents the state-of-the-art LLM adoption and adaptation in transportation, but also proposes future research directions as well as insights and recommendations for policymakers and practitioners, paving the way for greater LLM-driven research innovations in transportation in the future.

42 ENGINEERING↗
DOE OSTI2025

Self-adaptive weights based on balanced residual decay rate for physics-informed neural networks and deep operator networks

Physics-informed deep learning has emerged as a promising alternative for solving partial differential equations. However, for complex problems, training these networks can still be challenging, often resulting in unsatisfactory accuracy and efficiency. In this work, we demonstrate that the failure of plain physics-informed neural networks arises from the significant discrepancy in the convergence rate of residuals at different training points, where the slowest convergence rate dominates the overall solution convergence. Based on these observations, we propose a pointwise adaptive weighting method that balances the residual decay rate across different training points. The performance of our proposed adaptive weighting method is compared with current state-of-the-art adaptive weighting methods on benchmark problems for both physics-informed neural networks and physics-informed deep operator networks. In conclusion, through extensive numerical results we demonstrate that our proposed approach of balanced residual decay rates offers several advantages, including bounded weights, high prediction accuracy, fast convergence rate, low training uncertainty, low computational cost, and ease of hyperparameter tuning.

Balanced convergence rate↗
DOE OSTI2024

Application of the AI2 Climate Emulator to E3SMv2's Global Atmosphere Model, With a Focus on Precipitation Fidelity

Abstract Can the current successes of global machine learning‐based weather simulators be generalized beyond 2‐week forecasts to stable and accurate multiyear runs? The recently developed AI2 Climate Emulator (ACE) suggests this is feasible, based upon 10‐year simulations with a network trained on output from a physics‐based global atmosphere model using a grid spacing of approximately 110 km and forced by a repeating annual cycle of sea‐surface temperature. Here we show that ACE, without modification, can be trained to emulate another major atmospheric model, EAMv2, run at a comparable grid spacing for at least 10 years with similarly small climate biases—a prerequisite to wider applicability. With an analysis that combines multiple temporal, spatial, and frequency domain perspectives, we show that ACE faithfully represents the spatiotemporal structure of EAMv2 precipitation and related variables. Finally, we show that a pretrained ACE network is able to adapt to a new global climate model simulation data set with 10 fewer training steps than when starting from random initialization, all while still maintaining low levels of climate bias. Further analysis of these fine‐tuning experiments reveal ACE's intriguing ability to interpolate between distinct global climate models.

Duncan, James P. C.↗
DOE OSTI2023

High precision control of laser energy for laser–matter interaction studies

Precise, highly reproducible control of the laser energy is required for high confidence laser–matter interaction research such as in dynamic compression science and high energy density physics. The energy must be adjustable without affecting the pulse shape (time varying intensity) or beam smoothness. We have developed a convenient two-stage energy tuning method for a nominal 100 J, 351 nm (UV) laser. The energy is adjusted in 10 J (10%) increments by operating the laser at full energy and inserting a beam splitter in the laser output. As the splitter is located after the final frequency tripling optics, the UV pulse shape is unchanged. The energy is varied by substituting a splitter of different reflectivity. For finer 3 J (3%) increments, the infrared pulse is attenuated inside the laser before the final amplifier. This requires modest tuning to preserve the pulse shape. The demonstrated variation in shot-to-shot reproducibility is less than +/–2.5 J (5% of the full energy), irrespective of the laser output energy. These approaches can be adapted to most ~100 J class lasers. We describe these techniques and show two examples where they have elucidated the underlying physics in laser shock compression experiments. One used only the beam splitters to establish the pressure for melting in iron. The other combined both techniques to finely increment the peak stress (~2 GPa steps) in germanium to precisely determine the onset and completion of melting—including the melting kinetics. Furthermore, these unambiguous results would not be possible without the developments described here.

47 OTHER INSTRUMENTATION↗
DOE OSTI2023

KEBLM: Knowledge-Enhanced Biomedical Language Models

Pretrained language models (PLMs) have demonstrated strong performance on many natural language processing (NLP) tasks. Despite their great success, these PLMs are typically pretrained only on unstructured free texts without leveraging existing structured knowledge bases that are readily available for many domains, especially scientific domains. As a result, these PLMs may not achieve satisfactory performance on knowledge-intensive tasks such as biomedical NLP. Comprehending a complex biomedical document without domain-specific knowledge is challenging, even for humans. Inspired by this observation, we propose a general framework for incorporating various types of domain knowledge from multiple sources into biomedical PLMs. We encode domain knowledge using lightweight adapter modules, bottleneck feed-forward networks that are inserted into different locations of a backbone PLM. For each knowledge source of interest, we pretrain an adapter module to capture the knowledge in a self-supervised way. We design a wide range of self-supervised objectives to accommodate diverse types of knowledge, ranging from entity relations to description sentences. Once a set of pretrained adapters is available, we employ fusion layers to combine the knowledge encoded within these adapters for downstream tasks. Each fusion layer is a parameterized mixer of the available trained adapters that can identify and activate the most useful adapters for a given input. Our method diverges from prior work by including a knowledge consolidation phase, during which we teach the fusion layers to effectively combine knowledge from both the original PLM and newly-acquired external knowledge using a large collection of unannotated texts. After the consolidation phase, the complete knowledge-enhanced model can be fine-tuned for any downstream task of interest to achieve optimal performance. Extensive experiments on many biomedical NLP datasets show that our proposed framework consistently improves the performance of the underlying PLMs on various downstream tasks such as natural language inference, question answering, and entity linking. These results demonstrate the benefits of using multiple sources of external knowledge to enhance PLMs and the effectiveness of the framework for incorporating knowledge into PLMs. Finally, while primarily focused on the biomedical domain in this work, our framework is highly adaptable and can be easily applied to other domains, such as the bioenergy sector.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2025

Development of Automated Atom Probe Tomography capability to study the influence of applied voltage and laser power on the final apparent composition of the analyzed specimen

This study presents the development and implementation of an autonomous Bayesian optimization (BO) framework for controlling and optimizing experimental parameters in Atom Probe Tomography (APT). Using commercial silicon needle samples as a benchmark system, we demonstrate that BO can efficiently navigate the complex parameter space of voltage and laser power to achieve target charge state ratios (specifically Si + /(Si + +Si 2+ )) with minimal experimental evaluations. Our implementation integrates Gaussian Process modeling with the CAMECA atom probe control framework, enabling autonomous adjustment of experimental conditions in real-time. Results show that the algorithm successfully converges to target ratios under different scenarios: maintaining a reference ratio, increasing the ratio (favoring Si 1+ ), and decreasing the ratio (favoring Si 2+ ). The system adapts to specimen evolution during analysis, compensating for changes in apex geometry while maintaining optimization targets. This work establishes a proof of concept for AI-driven optimization in APT, addressing the traditional challenges of manual parameter tuning and paving the way for applications to more complex materials where compositional accuracy is critical.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Grid-Interactive Multi-Zone Building Control Using Reinforcement Learning with Global-Local Policy Search

In this paper, we develop a grid-interactive multi-zone building controller based on a deep reinforcement learning (RL) approach. The controller is designed to facilitate building operation during normal conditions and demand response events, while ensuring occupants comfort and energy efficiency. We leverage a continuous action space RL formulation, and devise a two-stage global-local RL training framework. In the first stage, a global fast policy search is performed using a gradient-free RL algorithm. In the second stage, a local fine-tuning is conducted using a policy gradient method. In contrast to the state-of-the-art model predictive control (MPC) approach, the proposed RL controller does not require complex computation during real-time operation and can adapt to nonlinear building models. We illustrate the controller performance numerically using a five-zone commercial building.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2021

Grid-Interactive Multi-Zone Building Control Using Reinforcement Learning with Global-Local Policy Search: Preprint

In this paper, we develop a grid-interactive multi-zone building controller based on a deep reinforcement learning (RL) approach. The controller is designed to facilitate building operation during normal conditions and demand response events, while ensuring occupants comfort and energy efficiency. We leverage a continuous action space RL formulation, and devise a two-stage global-local RL training framework. In the first stage, a global fast policy search is performed using a gradient-free RL algorithm. In the second stage, a local fine-tuning is conducted using a policy gradient method. In contrast to the state-of-the-art model predictive control (MPC) approach, the proposed RL controller does not require complex computation during real-time operation and can adapt to non-linear building models. We illustrate the controller performance numerically using a five-zone commercial building.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2022

Trigonal symmetry breaking and its electronic effects in the two-dimensional dihalides MX 2 and trihalides MX 3

We study the consequences of the approximately trigonal (D 3d ) point symmetry of the transition metal (M) site in two-dimensional van der Waals MX 2 dihalides and MX 3 trihalides. The trigonal symmetry leads to a 2-2-1 orbital splitting of the transition metal d shell, which is best represented by d orbitals that describe the trigonal rather than Oh symmetry. The ligand-ligand bond length differences (rather than metal-ligand) take the role of a Jahn-Teller-like mode and in combination with interlayer distances and dimensionality effects tune the crystal field splittings and bandwidths. These effects, in turn, are amplified by electronic correlation effects—leading to crystal field splittings of the order of 0.1–1 eV between the singlet and the lowest orbital doublet—as opposed to the often assumed degenerate 3 t 2g states. Our calculations explain why most of the materials in this family are insulating, and why these considerations have to be taken into account in realistic models of them. Further, orbital order coupled to various lower symmetry lattice modes may lift the remaining orbital degeneracies, and we explain how these may support unique electronic states using ZrI 2 and CuCl 2 as examples, and offer a brief overview of possible electronic configurations in this class of materials. By building and analyzing Wannier models adapted to the appropriate symmetry we examine how the interplay among trigonal symmetry, electronic correlation effects, and – d orbital charge transfer leads to insulating, orbitally polarized magnetic and/or orbital-selective Mott states, and we provide a simple framework and numerical tool for others. Our paper establishes a rigorous framework to understand, control, and tune the electronic states in low-dimensional correlated halides. Furthermore, our analysis shows that trigonal symmetry and its breaking is a key feature of the two-dimensional halides that needs to be accounted for in search of novel electronic states in materials ranging from CrI 3 to α–RuCl 3 .

2-dimensional systems↗
DOE OSTI2025

Constrained GAN-Generated X-Ray CT Data For Self-Supervised And Foundation-Model Segmentation Of Concrete Microstructures

Three-dimensional characterization of materials using X-ray computed tomography (XCT) is challenging due to the complexity of internal structures, noise, and variations in resolution. Traditional computer vision models often struggle to accurately segment these images, particularly in domain-specific applications like materials science. While supervised deep learning approaches have been developed to address the limitations of conventional algorithms, they typically require large amounts of labeled training data and often fail to generalize across different datasets. Self-supervised, few-and zero-shot learning methods have gained prominence in natural image processing and segmentation tasks, but their application to scientific imaging remains limited due to the unique structural complexity, noise, and textural artifacts present in materials science data. In this work, we investigate how domain adaptation, leveraging physics-based and GAN-generated synthetic data, impacts segmentation performance. We introduce a modified Contrastive Unpaired Translation (CUT) model designed to generate realistic labeled data, which can be used for training, pre-training, and fine-tuning segmentation models for real XCT microstructure data. We evaluate the performance of two segmentation approaches: a self-supervised network (SSL-ALPNet) and a foundation model (Segment Anything Model), assessing their improvements when pre-trained and/or fine-tuned on the synthesized data. Our results demonstrate that leveraging synthetic data significantly enhances segmentation performance, particularly in challenging materials science applications.

Ziabari, Amir [ORNL] (ORCID:000000034776457X)↗
DOE OSTI2023

GenSLMs: Genome-scale language models reveal SARS-CoV-2 evolutionary dynamics

We seek to transform how new and emergent variants of pandemic-causing viruses, specifically SARS-CoV-2, are identified and classified. By adapting large language models (LLMs) for genomic data, we build genome-scale language models (GenSLMs) which can learn the evolutionary landscape of SARS-CoV-2 genomes. By pre-training on over 110 million prokaryotic gene sequences and fine-tuning a SARS-CoV-2-specific model on 1.5 million genomes, we show that GenSLMs can accurately and rapidly identify variants of concern. Thus, to our knowledge, GenSLMs represents one of the first whole-genome scale foundation models which can generalize to other prediction tasks. We demonstrate scaling of GenSLMs on GPU-based supercomputers and AI-hardware accelerators utilizing 1.63 Zettaflops in training runs with a sustained performance of 121 PFLOPS in mixed precision and peak of 850 PFLOPS. We present initial scientific insights from examining GenSLMs in tracking evolutionary dynamics of SARS-CoV-2, paving the path to realizing this on large biological data.

Zvyagin, Maxim↗
DOE OSTI2026

MATEY: multiscale adaptive transformer models for spatiotemporal physical systems

Accurate representation of the multiscale features in spatiotemporal physical systems using vision transformer architectures requires extremely long, computationally prohibitive token sequences. To address this issue, we propose two novel adaptive tokenization schemes that dynamically adjust patch sizes based on local features: one ensures convergent behavior to uniform patch refinement, while the other offers better computational efficiency. Moreover, we present a set of spatiotemporal attention schemes, where the temporal or axial spatial dimensions are decoupled, to evaluate their baseline computational and data efficiencies and to determine whether adaptive tokenization can improve this performance. We assess the performance of the proposed multiscale adaptive model, MATEY, in a sequence of experiments. Compared to a full spatiotemporal attention scheme or a scheme that decouples only the temporal dimension, we find that fully decoupled axial attention is less efficient and expressive, requiring more training time and model parameters to achieve the same accuracy. The experiments on the adaptive tokenization schemes show that, compared to a uniformly refined model, the proposed schemes achieve comparable or improved accuracy at a much lower cost in the tested two-dimensional settings. While the asymptotic analysis suggests the potential for favorable scaling, empirical validation at substantially longer sequence lengths remains to be performed in future work. Finally, we demonstrate in two fine-tuning tasks featuring different physics that models pretrained on PDEBench data outperform the ones trained from scratch, especially in the low data regime with frozen attention.

adaptive tokenization↗
DOE OSTI2020

Safety evaluation of connected and automated vehicles in mixed traffic with conventional vehicles at intersections

Connected and Automated Vehicles (CAVs) can potentially improve the performance of the transportation system by reducing human errors. This paper investigates the safety impact of CAVs in a mixed traffic with conventional vehicles at intersections. Analyzing real-world AV crashes in California revealed that rear-end crashes at intersections are the dominant crash type. Therefore, to enhance our understanding of the future interactions between human-driven vehicles with CAVs at intersections, a simulation framework was developed to model the mixed traffic environment of Automated Vehicles (AV), cooperative AVs, and conventional human-driven vehicles. In order to model AVs driving behavior, Adaptive Cruise Control (ACC) and cooperative ACC (CACC) models are utilized. Particularly, this study explores system improvements due to automation and connectivity across varying CAV market penetration scenarios. ACC and CACC car following models are used to mimic the behavior of AVs and cooperative AVs. Real-world connected vehicle data are utilized to modify and tune the acceleration/deceleration regimes of the Wiedemann model. Next, the driving volatility concept capturing variability in vehicle speeds was utilized to calibrate the simulation to represent the safety performance of a real-world environment. Two surrogate safety measures are used to evaluate the safety performance of a representative intersection under different market penetration rate of CAVs: the number of longitudinal conflicts and driving volatility. At low levels of ACC market penetration, the safety improvements were found to be marginal, but safety improved substantially with more than 40% ACC penetration. Additional safety improvements can be achieved more quickly through the addition of cooperation and connectivity through CACC. Furthermore, ACC/CACC vehicles were found to improve mobility performance in terms of average speed and travel time at intersections.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2021

Measuring Saturn's Electron Beam Energy Spectrum using Webb's Wedges

It is very difficult to measure the voltage of the load on the Saturn accelerator. Time-resolved measurements such as vacuum voltmeters and V-dot monitors are impractical at best and completely change the pulsed power behavior at the load at worst. We would like to know the load voltage of the machine so that we could correctly model the radiation transport and tune our x-ray unfold methodology and circuit simulations of the accelerator. Step wedges have been used for decades as a tool to measure the end - point energies of high energy particle beams. Typically, the technique is used for multi-megavolt accelerators, but we have adapted it to Saturn's modest <2 MV end-point energy and modified the standard bremsstrahlung x-ray source to extract the electron beam without changing the physics of the load region. We found clear evidence of high energy electrons >2 MV. We also attempted to unfold an electron energy spectrum using a machine learning algorithm and while these results come with large uncertainties, they qualitatively agree with PIC simulation results.

43 PARTICLE ACCELERATORS↗
DOE OSTI2024

Tuning Exciton Emission via Ferroelectric Polarization at a Heterogeneous Interface between a Monolayer Transition Metal Dichalcogenide and a Perovskite Oxide Membrane

Here we demonstrate the integration of a thin BaTiO 3 (BTO) membrane with monolayer MoSe 2 in a dual gate device that enables in-situ manipulation of the BTO ferroelectric polarization with a voltage pulse. While two-dimensional (2D) transition metal dichalcogenides (TMDs) offer remarkable adaptability, their hybrid integration with other families of functional materials beyond the realm of 2D materials has been challenging. Released functional oxide membranes offer a solution for 2D/3D integration via stacking. 2D TMD excitons can serve as a local probe of the ferroelectric polarization in BTO at a heterogeneous interface. Using photoluminescence (PL) of MoSe 2 excitons to optically readout the doping level, we find that the relative population of charge carriers in MoSe 2 depends sensitively on the ferroelectric polarization. This finding points to a promising avenue for future-generations versatile sensing devices with high sensitivity, fast read-out, and diverse applicability for advanced signal processing.

42 ENGINEERING↗
DOE OSTI2025

A molecular conveyor belt-associated protein controls the rotational direction of the bacterial type 9 secretion system

ABSTRACT Many bacteria utilize the type 9 secretion system (T9SS) for gliding motility, surface colonization, and pathogenesis. This dual-function motor supports both gliding motility and protein secretion, where rotation of the T9SS plays a central role. Fueled by the energy of the stored proton motive force and transmitted through the torque of membrane-anchored stator units, the rotary T9SS propels an adhesin-coated conveyor belt along the bacterial outer membrane like a molecular snowmobile, thereby enabling gliding motion. However, the mechanisms controlling the rotational direction and gliding motility of T9SS remain elusive. Shedding light on this mechanism, we find that in the gliding bacteriumFlavobacterium johnsoniae, deletion of the C-terminus of the conveyor belt-associated protein GldJ controls and, in fact, reverses the rotational direction of T9SS from counterclockwise (CCW) to clockwise (CW). This suggests that the interface between the conveyor belt-associated protein GldJ and the T9SS ring protein GldK plays an important role in controlling the directionality of T9SS, potentially by modulating its interaction with the stator complex GldLM, which drives motor rotation. Combined with MD simulation of the T9SS stator units GldLM, we suggest a “tri-component gearset” model where GldJ controls the rotational direction of its driver, the T9SS, thus providing adaptive sensory feedback to influence the motility of the gliding bacterium. IMPORTANCE The type 9 secretion system (T9SS) is fundamental to bacterial gliding motility, pathogenesis, and surface colonization. Our findings reveal that the C-terminal region of the conveyor belt-associated protein GldJ functions as a molecular switch which is capable of reversing the rotational direction of T9SS. Through the coordinated actions of the T9SS stator units (akin to a driving motor), the GldK ring (the gear that converts rotational energy into linear movement), and GldJ, this machinery forms a smart conveyor belt system reminiscent of flexible or cognitive mechanical conveyors. Such advanced conveyors can alter their direction to adapt to shifting demands. Here, we show that the bacterial T9SS similarly adjusts its rotational bias based on feedback from the conveyor belt-associated protein GldJ. This dual-role feedback mechanism underscores an evolved, controllable biological snowmobile, offering new avenues for studying how bacteria fine-tune motility in dynamic environments.

Microbiology↗
DOE OSTI2026

CTRL-STEER: Closed-Loop Neuron Activation Control in Vision-Language-Action Models

Vision-Language-Action (VLA) models enable test-time behavioral steering via neuron-level interventions, but existing methods use fixed strengths and operate in open loop. This static modulation fails under evolving task dynamics, leading to overcorrection, oscillations, and reduced task success—especially for temporal attributes like speed. We propose CTRL-STEER, a control-theoretic framework that casts activation steering as closed-loop feedback with adaptive, time-varying interventions. Instead of assuming neurons encode temporal concepts, we steer along motion-aligned residual directions and regulate intervention magnitude via feedback. We instantiate this with both PID and reinforcement learning controllers that jointly optimize concept adherence and task success. Experiments on fine-tuned OpenVLA policies across four LIBERO suites show improved stability and a better steering–success trade-off over fixed-coefficient baselines, without retraining the base model.

Babu, Abhijith [Florida International University, ↗
DOE OSTI2025

SymbolNet: neural symbolic regression with adaptive dynamic pruning for compression

Abstract Compact symbolic expressions have been shown to be more efficient than neural network (NN) models in terms of resource consumption and inference speed when implemented on custom hardware such as field-programmable gate arrays (FPGAs), while maintaining comparable accuracy (Tsoi et al 2024 EPJ Web Conf. 295 09036). These capabilities are highly valuable in environments with stringent computational resource constraints, such as high-energy physics experiments at the CERN Large Hadron Collider. However, finding compact expressions for high-dimensional datasets remains challenging due to the inherent limitations of genetic programming (GP), the search algorithm of most symbolic regression (SR) methods. Contrary to GP, the NN approach to SR offers scalability to high-dimensional inputs and leverages gradient methods for faster equation searching. Common ways of constraining expression complexity often involve multistage pruning with fine-tuning, which can result in significant performance loss. In this work, we propose S y m b o l N e t , a NN approach to SR specifically designed as a model compression technique, aimed at enabling low-latency inference for high-dimensional inputs on custom hardware such as FPGAs. This framework allows dynamic pruning of model weights, input features, and mathematical operators in a single training process, where both training loss and expression complexity are optimized simultaneously. We introduce a sparsity regularization term for each pruning type, which can adaptively adjust its strength, leading to convergence at a target sparsity ratio. Unlike most existing SR methods that struggle with datasets containing more than O ( 10 ) inputs, we demonstrate the effectiveness of our model on the LHC jet tagging task (16 inputs), MNIST (784 inputs), and SVHN (3072 inputs).

Tsoi, Ho Fung (ORCID:0000000225502184)↗