Search NASA⌕ Search

SEARCH · Search NASA

Results for “Reasoning AI”

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
All sourcesNASA NTRSDOE OSTI
NASA NTRS2019

Artificial Intelligence: Powering Human Exploration of the Moon and Mars

Artificial Intelligence (AI) is a growing field of computa- tional science techniques designed to mimic functions per- formed by people. Advancements in autonomy will depend on a portfolio of AI technologies. Automated planning and scheduling is a venerable field of study in AI, and is needed for a variety of mission planning functions. Plan execution technology is less well studied, but important for auton- omy and robotics. Specialized forms of automated reason- ing and machine learning are key technologies to enable fault management. Over the past decade, the NASA Au- tonomous Systems and Operations (ASO) project has devel- oped and demonstrated numerous autonomy enabling tech- nologies employing AI techniques. Our work has employed AI in three distinct ways to enable autonomous mission op- erations capabilities. Crew Autonomy gives astronauts tools to assist in the performance of each of these mission oper-ations functions. Vehicle System Management uses AI tech- niques to turn the astronaut's spacecraft into a robot, allow- ing it to operate when astronauts are not present, or to reduce astronaut workload. AI technology also enables Autonomous Robots as crew assistants or proxies when the crew are not present. When these capabilities are used to enable astro- nauts to operate autonomously, they must be integrated with user interfaces, introducing numerous human factors con- siderations; when these capabilities are used to enable vehi- cle system management, they must be integrated with flight software, and run on embedded processors under the control of real-time operating systems.We first describe human spaceflight mission operations capabilities. The remainder of the paper will describe the ASO project, and the development and demonstration per- formed by ASO since 2011. We will describe the AI tech- niques behind each of these demonstrations, which include a variety of symbolic automated reasoning and machine learn- ing based approaches. Finally, we conclude with an assess- ment of future development needs for AI to enable NASA's future Exploration missions.

Mission Operations↗
NASA NTRS1990

A distributed agent architecture for real-time knowledge-based systems: Real-time expert systems project, phase 1

We propose a distributed agent architecture (DAA) that can support a variety of paradigms based on both traditional real-time computing and artificial intelligence. DAA consists of distributed agents that are classified into two categories: reactive and cognitive. Reactive agents can be implemented directly in Ada to meet hard real-time requirements and be deployed on on-board embedded processors. A traditional real-time computing methodology under consideration is the rate monotonic theory that can guarantee schedulability based on analytical methods. AI techniques under consideration for reactive agents are approximate or anytime reasoning that can be implemented using Bayesian belief networks as in Guardian. Cognitive agents are traditional expert systems that can be implemented in ART-Ada to meet soft real-time requirements. During the initial design of cognitive agents, it is critical to consider the migration path that would allow initial deployment on ground-based workstations with eventual deployment on on-board processors. ART-Ada technology enables this migration while Lisp-based technologies make it difficult if not impossible. In addition to reactive and cognitive agents, a meta-level agent would be needed to coordinate multiple agents and to provide meta-level control.

Lee, S. Daniel↗
DOE OSTI2025

Agentic AI vs ML-Based Autotuning: A Comparative Study for Loop Reordering Optimization

High Performance Computing (HPC) applications rely heavily on code optimizations to achieve good performance on modern CPU and GPU architectures. Traditional Machine Learning auto-tuning approaches have demonstrated success in exploring high-dimensional spaces, but they often require expensive compile-run evaluations and lack adaptability for large HPC applications. The recent advances in Large Language Models (LLMs) and Agentic AI systems raise intriguing questions about the potential of these approaches to address specific optimization methodologies. This work aims to answer an essential question for the HPC community: “How Agentic AI Systems Compare to Traditional ML Autotuning Techniques?” To address this question, we present a comparative analysis between a traditional ML-based optimization approach and an Agentic AI system, evaluating their respective capabilities and limitations for loop-level optimization. In addition, we introduced a new Agentic AI system named LoopGen-AI using three different Large Language Models: GPT-4.1, Claude 4.0, and Gemini 2.5. A key finding is that LoopGen-AI achieves competitive per-formance with only a few program runs, the reasoning logs from the agents revealed that their decisions rely heavily on the combination of semantic understanding of the target kernel with dynamic feedback from the environment, highlighting a promising new dimension in performance tuning. In contrast, ML-based autotuners focus on statistical exploration, and require orders of magnitude more runs to reach peak performance. Additionally, our analysis shows that prompt engineering, particularly using Persona + Context Manager patterns, significantly impacts the effectiveness of Agentic AI. Our results indicate that while Agentic AI systems are not yet a complete replacement for ML-based autotuners, it can effectively complement traditional methods.

Rosas, Miguel Romero↗
DOE OSTI2025

AI-Powered Knowledge Graphs for Neuromorphic and Energy-Efficient Computing

The surge in scientific literature obscures breakthroughs and hinders the discovery of new research paths. We propose an artificial intelligence (AI) powered framework using large language models (LLMs) and knowledge graphs (KGs) to automate parts of scientific discovery, focusing on energy-efficient AI circuits. Our hybrid approach combines LLMs, structured data, and ontology-based reasoning to construct a comprehensive knowledge graph that integrates insights across computational neuroscience, spiking neuron models, learning rules, architectural motifs, and neuromorphic device technologies. This multi-domain representation enables the generation of hypotheses that connect biological function with implementable, energy-efficient hardware architectures. Using KG embeddings and graph neural networks, the framework generates hypotheses for novel circuits, validates them through optimization on exascale HPC systems, and with tools like SuperNeuro and Fugu, the most promising designs will be prototyped in hardware. This open-source system aims to accelerate discoveries and bridging neuroscience with hardware innovation, drive collaboration, and unlock new opportunities in low-power AI computing.

Gautam, Ashish [ORNL]↗
DOE OSTI2025

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Large language models (LLMs) are increasingly adapted to downstream tasks via reinforcement learning (RL) methods like Group Relative Policy Optimization (GRPO), which often require thousands of rollouts to learn new tasks. We argue that the interpretable nature of language often provides a much richer learning medium for LLMs, compared to policy gradients derived from sparse, scalar rewards. To test this, we introduce GEPA (Genetic-Pareto), a prompt optimizer that thoroughly incorporates natural language reflection to learn high-level rules from trial and error. Given any AI system containing one or more LLM prompts, GEPA samples trajectories (e.g., reasoning, tool calls, and tool outputs) and reflects on them in natural language to diagnose problems, propose and test prompt updates, and combine complementary lessons from the Pareto frontier of its own attempts. As a result of GEPA's design, it can often turn even just a few rollouts into a large quality gain. Across six tasks, GEPA outperforms GRPO by 6% on average and by up to 20%, while using up to 35x fewer rollouts. GEPA also outperforms the leading prompt optimizer, MIPROv2, by over 10% (e.g., +12% accuracy on AIME-2025), and demonstrates promising results as an inference-time search strategy for code optimization. We release our code at https://github.com/gepa-ai/gepa.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

DOE Repository Metadata Profile (DRMP): A Metadata Framework for Advancing Interoperability and AI Readiness Across Scientific Repositories

The Department of Energy (DOE) funds a diverse and distributed ecosystem of repositories that steward scientific data, publications, and software across its research programs, user facilities, and national laboratories. While significant progress has been made in standardizing dataset-level metadata, the metadata describing repositories themselves (their identity, governance, access interfaces, policies, and technical capabilities) remains inconsistent and fragmented across DOE-funded systems. This variability limits discoverability, interoperability, automated validation, and AI-driven analysis, all of which are increasingly essential for modern scientific workflows. To address this gap, the DOE Data Curation Working Group (DCWG) developed the DOE Repository Metadata Profile (DRMP). The DRMP is a practical, community-driven framework that defines how repositories can describe themselves in a consistent, machine-actionable, and scalable manner. The DRMP is not a new metadata schema. Instead, it is a mapping profile and structured element set capturing the essential characteristics of DOE repositories. It harmonizes repository-level metadata across six widely adopted community schemas: RE3Data; DCAT-US v3; Schema.org; Dublin Core; DataCite 4.6; and PREMIS 3.0. This harmonization eliminates reinvention and enables interoperability within DOE and across the broader scientific ecosystem. A core objective of the DRMP is to reduce burden on repositories by allowing them to reuse their existing metadata through a Rosetta-style crosswalk rather than redesigning local implementations. The profile introduces a three-level conformance model that supports incremental adoption: • Level 1 – Minimum Viable Record (MVR): foundational identification elements required for workflows, project registration, and basic repository presence. • Level 2 – Interoperable: structured metadata enabling alignment with national and international discovery systems. • Level 3 – AI-Ready: enhanced provenance, policy transparency, fixity, semantic context, and capabilities that support automated reasoning, model training governance, and machine-assisted curation. To support implementation, the DRMP includes JSON Schema definitions, OpenAPI patterns, and MCP templates that allow repositories to publish machine-readable metadata directly within existing platforms. These resources are modular and lightweight, enabling adoption without major architectural change. Adopting the DRMP enables repositories to: • Enhance discoverability and interoperability by aligning identifiers, classifications, and descriptive elements across widely used schema standards. • Support federated discovery and cross-registration across DOE systems, Data.gov, and international catalogs. • Enable AI agents and workflow orchestration systems to interpret repository-level metadata within the American Science Cloud (AmSC) through Model Context Protocol (MCP)-based context publication. • Demonstrate alignment with DOE’s open science, stewardship, and FAIR data priorities. This guidance represents a community-driven step forward. Through voluntary adoption and continued feedback, the DRMP advances a cohesive, machine-actionable description of DOE repositories that supports FAIR data practices, preparing the infrastructure for AI-enabled research, and strengthening the discoverability and reuse of DOE’s scientific outputs.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
NASA NTRS1990

Attitude Determination and Control System (ADCS) and Maintenance and Diagnostic System (MDS): A maintenance and diagnostic system for Space Station Freedom

The Maintenance and Diagnostic System (MDS) that is being developed at Honeywell to enhance the Fault Detection Isolation and Recovery system (FDIR) for the Attitude Determination and Control System on Space Station Freedom is described. The MDS demonstrates ways that AI-based techniques can be used to improve the maintainability and safety of the Station by helping to resolve fault anomalies that cannot be fully determined by built-in-test, by providing predictive maintenance capabilities, and by providing expert maintenance assistance. The MDS will address the problems associated with reasoning about dynamic, continuous information versus only about static data, the concerns of porting software based on AI techniques to embedded targets, and the difficulties associated with real-time response. An initial prototype was built of the MDS. The prototype executes on Sun and IBM PS/2 hardware and is implemented in the Common Lisp; further work will evaluate its functionality and develop mechanisms to port the code to Ada.

Toms, David↗
NASA NTRS1994

Thunderstorm Hypothesis Reasoner

THOR is a knowledge-based system which incorporates techniques from signal processing, pattern recognition, and artificial intelligence (AI) in order to determine the boundary of small thunderstorms which develop and dissipate over the area encompassed by KSC and the Cape Canaveral Air Force Station. THOR interprets electric field mill data (derived from a network of electric field mills) by using heuristics and algorithms about thunderstorms that have been obtained from several domain specialists. THOR generates two forms of output: contour plots which visually describe the electric field activity over the network and a verbal interpretation of the activity. THOR uses signal processing and pattern recognition to detect signatures associated with noise or thunderstorm behavior in a near real time fashion from over 31 electrical field mills. THOR's AI component generates hypotheses identifying areas which are under a threat from storm activity, such as lightning. THOR runs on a VAX/VMS at the Kennedy Space Center. Its software is a coupling of C and FORTRAN programs, several signal processing packages, and an expert system development shell.

Mulvehill, Alice M.↗
DOE OSTI2025

ciemss/pyciemss

Causal and probabilistic reasoning with continuous time dynamical systems

Witty, Sam [Sorbus AI]↗
NASA NTRS1994

PalymSys (TM): An extended version of CLIPS for construction and reasoning using blackboards

This paper describes PalymSys(TM) -- an extended version of the CLIPS language that is designed to facilitate the implementation of blackboard systems. The paper first describes the general characteristics of blackboards and shows how a control blackboard architecture can be used by AI systems to examine their own behavior and adapt to real-time problem-solving situations by striking a balance between domain and control reasoning. The paper then describes the use of PalymSys in the development of a situation assessment subsystem for use aboard Army helicopters. This system performs real-time inferencing about the current battlefield situation using multiple domain blackboards as well as a control blackboard. A description of the control and domain blackboards and their implementation is presented. The paper also describes modifications made to the standard CLIPS 6.02 language in PalymSys(TM) 2.0. These include: (1) a dynamic Dempster-Shafer belief network whose structure is completely specifiable at run-time in the consequent of a PalymSys(TM) rule, (2) extension of the run command including a continuous run feature that enables the system to run even when the agenda is empty, and (3) a built-in communications link that uses shared memory to communicate with other independent processes.

Bryson, Travis↗
DOE OSTI2025

Connecting Minds: AI Use Cases to Bridge Power Systems and Large Language Models for Practical Applications

Recent advances in artificial intelligence (AI) and development of large language models (LLMs) present the opportunity to develop a new generation of power systems applications. In contrast with early power system AI applications based on structured numerical data, LLMs offer unique capabilities to perform logical reasoning using text documents, unstructured data, and application programming interface (API) calls to computational software. This paper seeks to bridge the knowledge gap between power systems engineers and LLM developers through a crosscutting explanation of use cases, characteristics, requirements, practical considerations from the perspectives of both LLM capabilities and industry needs. Specific focus is given to applications that can be realistically deployed by electric utilities. After introducing the architecture of LLMs and unique challenges of the power systems domain, this paper proposes twenty representative LLM applications grouped into categories of 1) power system operations, 2) asset management, 3) system planning and analytics, and 4) energy management and protection systems. Five use cases are presented within each category with descriptions of the motivation, objectives, approaches, example inputs / outputs, and benefits of each use case.

24 POWER TRANSMISSION AND DISTRIBUTION↗
NASA NTRS1983

A system for intelligent teleoperation research

The Automation Technology Branch of NASA Langley Research Center is developing a research capability in the field of artificial intelligence, particularly as applicable in teleoperator/robotics development for remote space operations. As a testbed for experimentation in these areas, a system concept has been developed and is being implemented. This system termed DAISIE (Distributed Artificially Intelligent System for Interacting with the Environment), interfaces the key processes of perception, reasoning, and manipulation by linking hardware sensors and manipulators to a modular artificial intelligence (AI) software system in a hierarchical control structure. Verification experiments have been performed: one experiment used a blocksworld database and planner embedded in the DAISIE system to intelligently manipulate a simple physical environment; the other experiment implemented a joint-space collision avoidance algorithm. Continued system development is planned.

Orlando, N. E.↗
NASA NTRS1988

Dynamic reasoning in a knowledge-based system

Any space based system, whether it is a robot arm assembling parts in space or an onboard system monitoring the space station, has to react to changes which cannot be foreseen. As a result, apart from having domain-specific knowledge as in current expert systems, a space based AI system should also have general principles of change. This paper presents a modal logic which can not only represent change but also reason with it. Three primitive operations, expansion, contraction and revision are introduced and axioms which specify how the knowledge base should change when the external world changes are also specified. Accordingly the notion of dynamic reasoning is introduced, which unlike the existing forms of reasoning, provide general principles of change. Dynamic reasoning is based on two main principles, namely minimize change and maximize coherence. A possible-world semantics which incorporates the above two principles is also discussed. The paper concludes by discussing how the dynamic reasoning system can be used to specify actions and hence form an integral part of an autonomous reasoning and planning system.

Rao, Anand S.↗
DOE OSTI2022

Workforce Development Bioenergy Experiential Learning Tool (Final Report)

The "Pathways to Bio-Power" program was designed as a workforce development initiative aimed at addressing the workforce gap in the bioenergy sector while promoting diversity and inclusion. The program incorporated advanced technologies, such as artificial intelligence (AI) and learning algorithms, to provide individualized training plans and pathways for Minority Business Enterprises (MBEs) and the current workforce to enter the bioenergy industry. However, after careful consideration, it was determined that proceeding to Phase II was not feasible due to several reasons related to resource allocation, market demand, and technical challenges.

15 GEOTHERMAL ENERGY↗
NASA NTRS1987

Geometric reasoning

Cognitive robot systems are ones in which sensing and representation occur, from which task plans and tactics are determined. Such a robot system accomplishes a task after being told what to do, but determines for itself how to do it. Cognition is required when the work environment is uncontrolled, when contingencies are prevalent, or when task complexity is large; it is useful in any robotic mission. A number of distinguishing features can be associated with cognitive robotics, and one emphasized here is the role of artificial intelligence in knowledge representation and in planning. While space telerobotics may elude some of the problems driving cognitive robotics, it shares many of the same demands, and it can be assumed that capabilities developed for cognitive robotics can be employed advantageously for telerobotics in general. The top level problem is task planning, and it is appropriate to introduce a hierarchical view of control. Presented with certain mission objectives, the system must generate plans (typically) at the strategic, tactical, and reflexive levels. The structure by which knowledge is used to construct and update these plans endows the system with its cognitive attributes, and with the ability to deal with contingencies, changes, unknowns, and so on. Issues of representation and reasoning which are absolutely fundamental to robot manipulation, decisions based upon geometry, are discussed here, not AI task planning per se.

Woodbury, R. F.↗
DOE OSTI2021

Artificial Intelligence for Autonomous Molecular Design: A Perspective

Domain-aware artificial intelligence has been increasingly adopted in recent years to expedite molecular design in various applications, including drug design and discovery. Recent advances in areas such as physics-informed machine learning and reasoning, software engineering, high-end hardware development, and computing infrastructures are providing opportunities to build scalable and explainable AI molecular discovery systems. This could improve a design hypothesis through feedback analysis, data integration that can provide a basis for the introduction of end-to-end automation for compound discovery and optimization, and enable more intelligent searches of chemical space. Several state-of-the-art ML architectures are predominantly and independently used for predicting the properties of small molecules, their high throughput synthesis, and screening, iteratively identifying and optimizing lead therapeutic candidates. However, such deep learning and ML approaches also raise considerable conceptual, technical, scalability, and end-to-end error quantification challenges, as well as skepticism about the current AI hype to build automated tools. To this end, synergistically and intelligently using these individual components along with robust quantum physics-based molecular representation and data generation tools in a closed-loop holds enormous promise for accelerated therapeutic design to critically analyze the opportunities and challenges for their more widespread application. This article aims to identify the most recent technology and breakthrough achieved by each of the components and discusses how such autonomous AI and ML workflows can be integrated to radically accelerate the protein target or disease model-based probe design that can be iteratively validated experimentally. Taken together, this could significantly reduce the timeline for end-to-end therapeutic discovery and optimization upon the arrival of any novel zoonotic transmission event. Our article serves as a guide for medicinal, computational chemistry and biology, analytical chemistry, and the ML community to practice autonomous molecular design in precision medicine and drug discovery.

59 BASIC BIOLOGICAL SCIENCES↗
NASA NTRSDate not supplied

Artificial Intelligence Enhancements to Imagery for Space Operations

Philosophy classes still ponder the question asked by Dr. George Berkely, an Anglican Bishop and philosopher in the 1600’s-- “If a tree falls in a forest and no one is around to hear it, does it make a sound?” With that in mind, I ask the following—If a still image or motion imagery from a space mission cannot be found during a search, does it exist? Since the beginning of spaceflight, imagery has been a key form of data collected. Whether for mere curiosity (what does Earth look like from Space?), or for operational reasons (did the solar panel deploy?), or for engineering purposes (what was that object that floated away from the spacecraft?), imagery has been included in space missions. To be useful, though, the image or motion imagery must be accessible and accessed when needed. During the analog era, that typically meant captions and numbers associated with the physical media. With “born digital” imagery, it is possible to add metadata to the image data file. This metadata might include the date and time of capture, mission, camera, exposure data, and similar data fields. Many modern cameras embed some basic metadata into the image file at the moment of capture. The reality, though, is even with today’s born-digital enhancements with embedded metadata at the time of capture, reviewing and cataloging still and motion imagery is very labor intensive. Humans review the imagery for sensitive content (privacy concerns, imagery containing proprietary data/subject matter), and to identify imagery containing crew members or imagery that should be reviewed for engineering or scientific reasons. All this review and manual data entry is very time-consuming. Many improvements in Artificial Intelligence (AI), Machine Learning, and processing power now make it possible to identify persons, objects, motion, color, audio with sensitive content, and other details after or while the imagery is captured.

Rodney Grubbs↗
NASA NTRS2019

A New Regional Climate Model for POLAR‐CORDEX: Evaluation of a 30‐Year Hindcast with COSMO‐CLM^2 Over Antarctica

Continent‐wide climate information over the Antarctic Ice Sheet (AIS) is important to obtain accurate information of present climate and reduce uncertainties of the ice sheet mass balance response and resulting global sea level rise to future climate change. In this study, the COSMO‐CLM^2 Regional Climate Model is applied over the AIS and adapted for the specific meteorological and climatological conditions of the region. A 30‐year hindcast was performed and evaluated against observational records consisting of long‐term ground‐based meteorological observations, automatic weather stations, radiosoundings, satellite records, stake measurements and ice cores. Reasonable agreement regarding the surface and upper‐air climate is achieved by the COSMO‐CLM^2 model, comparable to the performance of other state‐of‐the‐art climate models over the AIS. Meteorological variability of the surface climate is adequately simulated, and biases in the radiation and surface mass balance are small. The presented model therefore contributes as a new member to the COordinated Regional Downscaling EXperiment project over the AIS (POLAR‐CORDEX) and the CORDEX‐CORE initiative.

Antarctica↗