Search NASA⌕ Search

SEARCH · Search NASA

Results for “reasoning”

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 307 records · Page 17
All sourcesNASA NTRSDOE OSTI
NASA NTRS2012

Intelligent Data Visualization for Cross-Checking Spacecraft System Diagnosis

Any reasoning system is fallible, so crew members and flight controllers must be able to cross-check automated diagnoses of spacecraft or habitat problems by considering alternate diagnoses and analyzing related evidence. Cross-checking improves diagnostic accuracy because people can apply information processing heuristics, pattern recognition techniques, and reasoning methods that the automated diagnostic system may not possess. Over time, cross-checking also enables crew members to become comfortable with how the diagnostic reasoning system performs, so the system can earn the crew s trust. We developed intelligent data visualization software that helps users cross-check automated diagnoses of system faults more effectively. The user interface displays scrollable arrays of timelines and time-series graphs, which are tightly integrated with an interactive, color-coded system schematic to show important spatial-temporal data patterns. Signal processing and rule-based diagnostic reasoning automatically identify alternate hypotheses and data patterns that support or rebut the original and alternate diagnoses. A color-coded matrix display summarizes the supporting or rebutting evidence for each diagnosis, and a drill-down capability enables crew members to quickly view graphs and timelines of the underlying data. This system demonstrates that modest amounts of diagnostic reasoning, combined with interactive, information-dense data visualizations, can accelerate system diagnosis and cross-checking.

Ong, James C.↗
NASA NTRSDate not supplied

A Temporal Differential Dynamic Logic Formal Embedding

Differential dynamic logic is a formal framework to specify and reason about hybrid programs (HPs). The core of dL is a proof calculus that contains a collection of axioms and rules for the rigorous verification of properties of HPs. Recently, dL has been embedded within the theorem prover Prototype Verification System (PVS) resulting in the tool Plaidypvs2. The integration of dL into PVS expands its expressive power; user defined functions, such as trigonometric and other transcendental functions, can be used inside the dL framework, and meta-reasoning about HPs can be performed, including reasoning about entire classes of HPs, specified using dependent types in PVS. The differential temporal dynamic logic (dTL2) extends dL with temporal logic operators to reason about all the states reachable during the execution of an HP. This paper presents a work in progress focusing on embedding dTL2 in PVS as an extension of Plaidypvs. Plaidypvs is expanded with the formalization of a trace semantics for HPs, the definition of the LTL temporal operators eventually and globally, and the implementation of the proof calculus for dTL2. This new embedding has the same capabilties as Plaidypvs, which allows user defined functions and meta-reasoning of properties of HPs. To the best of the authors’ knowledge this is the first implementation of dTL2.

differential dynamic logic↗
DOE OSTI2024

On the applicability of the Redlich-Kister framework for viscosity estimation of molten halide salt mixtures

For molten halide salt mixtures already being utilized or under consideration for carbon-free energy production systems, it is crucial that their viscosity is well understood so that system thermal hydraulics can be reliably assessed. Because of the difficulty in accurately measuring molten halide viscosity and the sheer size of the matrix of possible higher order salt mixtures that may be of interest to the energy industry, there are several gaps in the quantified understanding of molten halide viscosity across this matrix. As such, both first-principles and semi-empirical modeling techniques may be crucial for rapidly assessing this broad, complex compositional domain. Herein, the Redlich-Kister framework is applied to assess the feasibility of broadly interpolating and estimating the viscosity of several pseudobinary and pseudoternary molten halide salt systems that may be of key interest to the energy industry. The framework is based on the assumption that an ideal component and a nonideal component collectively describe the viscosity as a function of composition and temperature for a given molten halide system. Three different ideal models were considered for the ideal component, including Grunburg-Nissan, Katti-Chaudhri, and Gambill methods. Regarding the pseudobinary interpolations, the Redlich-Kister models with either the Grunburg-Nissan or Katti-Chaudhri models as the ideal component resulted in either highly (average error less than 5%) or reasonably (average error less than 15%) accurate interpolations of pseudobinary halide viscosity; BeF 2 - or UF 4 -bearing salts tended to result in reasonably accurate interpolations, whereas other pseudobinary mixtures tended to show high accuracy. Regarding the pseudoternary extrapolations, the Redlich-Kister framework shows reasonable success at estimating the extent to which a pseudoternary system may indicate deviations from ideal Grunburg-Nissan mixing, where discrepancies with comparative experimental data generally stay within 30%. Finally, the primary reasons identified for such discrepancies are (1) inaccuracy in the underlying experimental data, (2) different complexation behavior in the higher order systems compared to the pseudobinary subsystems, and (3) extrapolation into temperatures too far out of the domain, which is valid for the underlying experimental data feeding the Redlich-Kister model.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1988

Transformation based endorsement systems

Evidential reasoning techniques classically represent support for a hypothesis by a numeric value or an evidential interval. The combination of support is performed by an arithmetic rule which often requires restrictions to be placed on the set of possibilities. These assumptions usually require the hypotheses to be exhausitive and mutually exclusive. Endorsement based classification systems represent support for the alternatives symbolically rather than numerically. A framework for constructing endorsement systems is presented in which transformations are defined to generate and update the knowledge base. The interaction of the knowledge base and transformations produces a non-monotonic reasoning system. Two endorsement based reasoning systems are presented to demonstrate the flexibility of the transformational approach for reasoning with ambiguous and inconsistent information.

Sudkamp, Thomas↗
NASA NTRS1990

StarPlan: A model-based diagnostic system for spacecraft

The Sunnyvale Division of Ford Aerospace created a model-based reasoning capability for diagnosing faults in space systems. The approach employs reasoning about a model of the domain (as it is designed to operate) to explain differences between expected and actual telemetry; i.e., to identify the root cause of the discrepancy (at an appropriate level of detail) and determine necessary corrective action. A development environment, named Paragon, was implemented to support both model-building and reasoning. The major benefit of the model-based approach is the capability for the intelligent system to handle faults that were not anticipated by a human expert. The feasibility of this approach for diagnosing problems in a spacecraft was demonstrated in a prototype system, named StarPlan. Reasoning modules within StarPlan detect anomalous telemetry, establish goals for returning the telemetry to nominal values, and create a command plan for attaining the goals. Before commands are implemented, their effects are simulated to assure convergence toward the goal. After the commands are issued, the telemetry is monitored to assure that the plan is successful. These features of StarPlan, along with associated concerns, issues and future directions, are discussed.

Heher, Dennis↗
NASA NTRS1990

Developing a framework for qualitative engineering: Research in design and analysis of complex structural systems

The research is focused on automating the evaluation of complex structural systems, whether for the design of a new system or the analysis of an existing one, by developing new structural analysis techniques based on qualitative reasoning. The problem is to identify and better understand: (1) the requirements for the automation of design, and (2) the qualitative reasoning associated with the conceptual development of a complex system. The long-term objective is to develop an integrated design-risk assessment environment for the evaluation of complex structural systems. The scope of this short presentation is to describe the design and cognition components of the research. Design has received special attention in cognitive science because it is now identified as a problem solving activity that is different from other information processing tasks (1). Before an attempt can be made to automate design, a thorough understanding of the underlying design theory and methodology is needed, since the design process is, in many cases, multi-disciplinary, complex in size and motivation, and uses various reasoning processes involving different kinds of knowledge in ways which vary from one context to another. The objective is to unify all the various types of knowledge under one framework of cognition. This presentation focuses on the cognitive science framework that we are using to represent the knowledge aspects associated with the human mind's abstraction abilities and how we apply it to the engineering knowledge and engineering reasoning in design.

Franck, Bruno M.↗
NASA NTRS1993

Driving a car with custom-designed fuzzy inferencing VLSI chips and boards

Vehicle control in a-priori unknown, unpredictable, and dynamic environments requires many calculational and reasoning schemes to operate on the basis of very imprecise, incomplete, or unreliable data. For such systems, in which all the uncertainties can not be engineered away, approximate reasoning may provide an alternative to the complexity and computational requirements of conventional uncertainty analysis and propagation techniques. Two types of computer boards including custom-designed VLSI chips were developed to add a fuzzy inferencing capability to real-time control systems. All inferencing rules on a chip are processed in parallel, allowing execution of the entire rule base in about 30 microseconds, and therefore, making control of 'reflex-type' of motions envisionable. The use of these boards and the approach using superposition of elemental sensor-based behaviors for the development of qualitative reasoning schemes emulating human-like navigation in a-priori unknown environments are first discussed. Then how the human-like navigation scheme implemented on one of the qualitative inferencing boards was installed on a test-bed platform to investigate two control modes for driving a car in a-priori unknown environments on the basis of sparse and imprecise sensor data is described. In the first mode, the car navigates fully autonomously, while in the second mode, the system acts as a driver's aid providing the driver with linguistic (fuzzy) commands to turn left or right and speed up or slow down depending on the obstacles perceived by the sensors. Experiments with both modes of control are described in which the system uses only three acoustic range (sonar) sensor channels to perceive the environment. Simulation results as well as indoors and outdoors experiments are presented and discussed to illustrate the feasibility and robustness of autonomous navigation and/or safety enhancing driver's aid using the new fuzzy inferencing hardware system and some human-like reasoning schemes which may include as little as six elemental behaviors embodied in fourteen qualitative rules.

Pin, Francois G.↗
NASA NTRS1993

Case-based approach to handling aircraft malfunctions

Reasoning about physical systems in operation is a difficult task, and any attempt to automate the process must overcome the problems of modeling normal behavior, diagnosing faults, and predicting future behavior. This paper describes a prototypical case-based reasoner (CBR) that operates in the domain of in-flight fault diagnosis and prognosis of aviation subsystems, particularly jet engines. The reasoner operates on the observation that the ability of a CBR program to reason about physical systems can be significantly enhanced by the addition to the CBR program of a model of the physical system to describe the system's structural, functional, and causal behavior.

Karamouzis, Stamos T.↗
NASA NTRS1993

Task planning with uncertainty for robotic systems

In a practical robotic system, it is important to represent and plan sequences of operations and to be able to choose an efficient sequence from them for a specific task. During the generation and execution of task plans, different kinds of uncertainty may occur and erroneous states need to be handled to ensure the efficiency and reliability of the system. An approach to task representation, planning, and error recovery for robotic systems is demonstrated. Our approach to task planning is based on an AND/OR net representation, which is then mapped to a Petri net representation of all feasible geometric states and associated feasibility criteria for net transitions. Task decomposition of robotic assembly plans based on this representation is performed on the Petri net for robotic assembly tasks, and the inheritance of properties of liveness, safeness, and reversibility at all levels of decomposition are explored. This approach provides a framework for robust execution of tasks through the properties of traceability and viability. Uncertainty in robotic systems are modeled by local fuzzy variables, fuzzy marking variables, and global fuzzy variables which are incorporated in fuzzy Petri nets. Analysis of properties and reasoning about uncertainty are investigated using fuzzy reasoning structures built into the net. Two applications of fuzzy Petri nets, robot task sequence planning and sensor-based error recovery, are explored. In the first application, the search space for feasible and complete task sequences with correct precedence relationships is reduced via the use of global fuzzy variables in reasoning about subgoals. In the second application, sensory verification operations are modeled by mutually exclusive transitions to reason about local and global fuzzy variables on-line and automatically select a retry or an alternative error recovery sequence when errors occur. Task sequencing and task execution with error recovery capability for one and multiple soft components in robotic systems are investigated.

Cao, Tiehua↗
NASA NTRS1995

Real-time value-driven diagnosis

Diagnosis is often thought of as an isolated task in theoretical reasoning (reasoning with the goal of updating our beliefs about the world). We present a decision-theoretic interpretation of diagnosis as a task in practical reasoning (reasoning with the goal of acting in the world), and sketch components of our approach to this task. These components include an abstract problem description, a decision-theoretic model of the basic task, a set of inference methods suitable for evaluating the decision representation in real-time, and a control architecture to provide the needed continuing coordination between the agent and its environment. A principal contribution of this work is the representation and inference methods we have developed, which extend previously available probabilistic inference methods and narrow, somewhat, the gap between probabilistic and logical models of diagnosis.

Dambrosio, Bruce↗
NASA NTRS1998

Transition in Turbine Flows

We have further developed our capabilities to analyze transition in turbine boundary layers from first principles by integrating the nonlinear parabolized stability equations (PSE) with improved initial and boundary conditions. With modified iteration schemes, we are able to proceed deeper into the transition region where skin friction coefficient and heat transfer coefficient significantly increase. Initial and boundary conditions at elevated turbulence levels can be derived by receptivity analysis. Test runs for ERCOFTAC test case T3A at 2.4\% turbulence level provide results in good agreement with the experimental data. The sharper minimum of the skin coefficient also shown by DNS results is likely due to the missing intermittency. The method has been applied to various experimentally studied turbine blades (UTRC, VKI, Zierke, Langston, Hippensteele, and others). The PSE results, though physically reasonable, do not agree as well as expected with the experimental findings. We have, therefore, performed an extensive search for the reasons of the seemingly systematic deviations. A first source of uncertainty has been found in the often insufficient documentation of the experiments (e.g. on blockage by end-wall boundary layers). However, variation of the relevant parameters does not lead to more satisfactory agreement. A second reason has been found in the "standard procedure" which considers a 2D flow at midspan and uses a panel code and subsequent boundary-layer code to obtain the laminar basic flow for the transition analysis. Comparison with the pressure distribution obtained with a 3D design code (RVC3D) shows significant three-dimensionality of the flow (e.g. in the UTRC experiments). The spanwise variation has been neglected in our original PSE code. To overcome this problem, we have developed the PSE/3D for fully 3D boundary layers to account for streamwise and spanwise variations. Since the design code does not provide the boundary-layer flow with sufficient resolution, we have generated the Euler solution and employed a 3D boundary-layer code to obtain the viscous basic flow. Although only the linear stability level of PSE/3D has been implemented so far, the discrepancies with the experiments change but do not disappear. We still find deviations between the computed and experimental variations of C(sub f), and St along the blade for laminar flow. The main reason can be seen by comparing the solution of the boundary-layer code with the viscous results of the design code. The conventional boundary-layer solution exhibits an asymptotic behavior appropriate in external aerodynamics but does not match the steep gradients of the inviscid flow through the passage and consequently provides biased results for C(sub f), and St. An attempt is currently being made to correct this deficiency. Before attempting to perform the transition analysis for the viscous flow provided by the design code, we have analyzed the implementation and "best possible" results. Code and results exhibit flaws that may negatively affect the design and are intolerable for transition analysis. Therefore, we have decided to develop a new code to obtain a reliable basis for stability and transition studies. We expect to report improved results by the time of the meeting.

Herbert, Thorwald↗
NASA NTRS2000

Autonomy Architectures for a Constellation of Spacecraft

Until the past few years, missions typically involved fairly large expensive spacecraft. Such missions have primarily favored using older proven technologies over more recently developed ones, and humans controlled spacecraft by manually generating detailed command sequences with low-level tools and then transmitting the sequences for subsequent execution on a spacecraft controller. This approach toward controlling a spacecraft has worked spectacularly on previous missions, but it has limitations deriving from communications restrictions - scheduling time to communicate with a particular spacecraft involves competing with other projects due to the limited number of deep space network antennae. This implies that a spacecraft can spend a long time just waiting whenever a command sequence fails. This is one reason why the New Millennium program has an objective to migrate parts of mission control tasks onboard a spacecraft to reduce wait time by making spacecraft more robust. The migrated software is called a "remote agent" and has 4 components: a mission manager to generate the high level goals, a planner/scheduler to turn goals into activities while reasoning about future expected situations, an executive/diagnostics engine to initiate and maintain activities while interpreting sensed events by reasoning about past and present situations, and a conventional real-time subsystem to interface with the spacecraft to implement an activity's primitive actions. In addition to needing remote planning and execution for isolated spacecraft, a trend toward multiple-spacecraft missions points to the need for remote distributed planning and execution. The past few years have seen missions with growing numbers of probes. Pathfinder has its rover (Sojourner), Cassini has its lander (Huygens), and the New Millenium Deep Space 3 (DS3) proposal involves a constellation of 3 spacecraft for interferometric mapping. This trend is expected to continue to progressively larger fleets. For example, one mission proposed to succeed DS3 would have 18 spacecraft flying in formation in order to detect earth-sized planets orbiting other stars. A proposed magnetospheric constellation would involve 5 to 500 spacecraft in Earth orbit to measure global phenomena within the magnetosphere. This work describes and compares three autonomy architectures for a system that continuously plans to control a fleet of spacecraft using collective mission goals instead of goals or command sequences for each spacecraft. A fleet of self-commanding spacecraft would autonomously coordinate itself to satisfy high level science and engineering goals in a changing partially-understood environment making feasible the operation of tens or even a hundred spacecraft (such as for interferometry or plasma physics missions). The easiest way to adapt autonomous spacecraft research to controlling constellations involves treating the constellation as a single spacecraft. Here one spacecraft directly controls the others as if they were connected. The controlling "master" spacecraft performs all autonomy reasoning, and the slaves only have real-time subsystems to execute the master's commands and transmit local telemetry/observations. The executive/diagnostics module starts actions and the master's real-time subsystem controls the action either locally or remotely through a slave. While the master/slave approach benefits from conceptual simplicity, it relies on an assumption that the master spacecraft's executive can continuously monitor the slaves' real-time subsystems, and this relies on high-bandwidth highly-reliable communications. Since unintended results occur fairly rarely, one way to relax the bandwidth requirements involves only monitoring unexpected events in spacecraft. Unfortunately, this disables the ability to monitor for unexpected events between spacecraft and leads to a host of coordination problems among the slaves. Also, failures in the communications system can result in losing slaves. The other two architectures improve robustness while reducing communications by progressively distributing more of the other three remote agent components across the constellation. In a teamwork architecture, all spacecraft have executives and real-time subsystems - only the leader has the planner/scheduler and mission manager. Finally, distributing all remote agent components leads to a peer-to-peer approach toward constellation control.

Barrett, Anthony↗
NASA NTRS1998

Model Based Autonomy for Robust Mars Operations

Space missions have historically relied upon a large ground staff, numbering in the hundreds for complex missions, to maintain routine operations. When an anomaly occurs, this small army of engineers attempts to identify and work around the problem. A piloted Mars mission, with its multiyear duration, cost pressures, half-hour communication delays and two-week blackouts cannot be closely controlled by a battalion of engineers on Earth. Flight crew involvement in routine system operations must also be minimized to maximize science return. It also may be unrealistic to require the crew have the expertise in each mission subsystem needed to diagnose a system failure and effect a timely repair, as engineers did for Apollo 13. Enter model-based autonomy, which allows complex systems to autonomously maintain operation despite failures or anomalous conditions, contributing to safe, robust, and minimally supervised operation of spacecraft, life support, In Situ Resource Utilization (ISRU) and power systems. Autonomous reasoning is central to the approach. A reasoning algorithm uses a logical or mathematical model of a system to infer how to operate the system, diagnose failures and generate appropriate behavior to repair or reconfigure the system in response. The 'plug and play' nature of the models enables low cost development of autonomy for multiple platforms. Declarative, reusable models capture relevant aspects of the behavior of simple devices (e.g. valves or thrusters). Reasoning algorithms combine device models to create a model of the system-wide interactions and behavior of a complex, unique artifact such as a spacecraft. Rather than requiring engineers to all possible interactions and failures at design time or perform analysis during the mission, the reasoning engine generates the appropriate response to the current situation, taking into account its system-wide knowledge, the current state, and even sensor failures or unexpected behavior.

Kurien, James A.↗
NASA NTRS2005

Visualization Methods for Viability Studies of Inspection Modules for the Space Shuttle

An effective simulation of an object, process, or task must be similar to that object, process, or task. A simulation could consist of a physical device, a set of mathematical equations, a computer program, a person, or some combination of these. There are many reasons for the use of simulators. Although some of the reasons are unique to a specific situation, there are many general reasons and purposes for using simulators. Some are listed but not limited to (1) Safety, (2) Scarce resources, (3) Teaching/education, (4) Additional capabilities, (5) Flexibility and (6) Cost. Robot simulators are in use for all of these reasons. Virtual environments such as simulators will eliminate physical contact with humans and hence will increase the safety of work environment. Corporations with limited funding and resources may utilize simulators to accomplish their goals while saving manpower and money. A computer simulation is safer than working with a real robot. Robots are typically a scarce resource. Schools typically don t have a large number of robots, if any. Factories don t want the robots not performing useful work unless absolutely necessary. Robot simulators are useful in teaching robotics. A simulator gives a student hands-on experience, if only with a simulator. The simulator is more flexible. A user can quickly change the robot configuration, workcell, or even replace the robot with a different one altogether. In order to be useful, a robot simulator must create a model that accurately performs like the real robot. A powerful simulator is usually thought of as a combination of a CAD package with simulation capabilities. Computer Aided Design (CAD) techniques are used extensively by engineers in virtually all areas of engineering. Parts are designed interactively aided by the graphical display of both wireframe and more realistic shaded renderings. Once a part s dimensions have been specified to the CAD package, designers can view the part from any direction to examine how it will look and perform in relation to other parts. If changes are deemed necessary, the designer can easily make the changes and view the results graphically. However, a complex process of moving parts intended for operation in a complex environment can only be fully understood through the process of animated graphical simulation. A CAD package with simulation capabilities allows the designer to develop geometrical models of the process being designed, as well as the environment in which the process will be used, and then test the process in graphical animation much as the actual physical system would be run . By being able to operate the system of moving and stationary parts, the designer is able to see in simulation how the system will perform under a wide variety of conditions. If, for example, undesired collisions occur between parts of the system, design changes can be easily made without the expense or potential danger of testing the physical system.

Mobasher, Amir A.↗
NASA NTRS2005

A Review of Diagnostic Techniques for ISHM Applications

System diagnosis is an integral part of any Integrated System Health Management application. Diagnostic applications make use of system information from the design phase, such as safety and mission assurance analysis, failure modes and effects analysis, hazards analysis, functional models, fault propagation models, and testability analysis. In modern process control and equipment monitoring systems, topological and analytic , models of the nominal system, derived from design documents, are also employed for fault isolation and identification. Depending on the complexity of the monitored signals from the physical system, diagnostic applications may involve straightforward trending and feature extraction techniques to retrieve the parameters of importance from the sensor streams. They also may involve very complex analysis routines, such as signal processing, learning or classification methods to derive the parameters of importance to diagnosis. The process that is used to diagnose anomalous conditions from monitored system signals varies widely across the different approaches to system diagnosis. Rule-based expert systems, case-based reasoning systems, model-based reasoning systems, learning systems, and probabilistic reasoning systems are examples of the many diverse approaches ta diagnostic reasoning. Many engineering disciplines have specific approaches to modeling, monitoring and diagnosing anomalous conditions. Therefore, there is no "one-size-fits-all" approach to building diagnostic and health monitoring capabilities for a system. For instance, the conventional approaches to diagnosing failures in rotorcraft applications are very different from those used in communications systems. Further, online and offline automated diagnostic applications are integrated into an operations framework with flight crews, flight controllers and maintenance teams. While the emphasis of this paper is automation of health management functions, striking the correct balance between automated and human-performed tasks is a vital concern.

Patterson-Hine, Ann↗
DOE OSTI2025

Can Large Language Models Understand Intermediate Representations?

Intermediate Representations (IRs) are essential in compiler design and program analysis, yet their comprehension by Large Language Models (LLMs) remains underexplored. This paper presents a pioneering empirical study to investigate the capabilities of LLMs, including GPT-4, GPT-3, Gemma 2, LLaMA 3.1, and Code Llama, in understanding IRs. We analyze their performance across four tasks: Control Flow Graph (CFG) reconstruction, decompilation, code summarization, and execution reasoning. Our results indicate that while LLMs demonstrate competence in parsing IR syntax and recognizing high-level structures, they struggle with control flow reasoning, execution semantics, and loop handling. Specifically, they often misinterpret branching instructions, omit critical IR operations, and rely on heuristic-based reasoning, leading to errors in CFG reconstruction, IR decompilation, and execution reasoning. The study underscores the necessity for IR-specific enhancements in LLMs, recommending fine-tuning on structured IR datasets and integration of explicit control flow models to augment their comprehension and handling of IR-related tasks.

Jiang, Hailong↗
NASA NTRSDate not supplied

Framing Potential Wildfire Opportunities for DRF

A Data & Reasoning Fabric (DRF) is envisioned to enable the full potential of air mobility by providing all data and reasoning where they are needed. The DRF provides a marketplace based on an open foundational ecosystem of data and reasoning exchange between the many systems that must seamlessly interplay to manage the complex and dense airspace operations required to achieve advanced air mobility goals. The DRF marketplace is decentralized and will not be owned by any single party. To deliver reference designs and development environments that catalyze broad private and public sector buy-in and self-sustaining development of DRF and associated standards, DRF activities will collaboratively test these technologies, open standards and architectures, and an integrated framework with end-users.

Aeronautics↗
DOE OSTI2026

AutoLabs: cognitive multi-agent systems with self-correction for autonomous chemical experimentation

The automation of chemical research through self-driving laboratories (SDLs) promises to accelerate scientific discovery, yet the reliability and granular performance of the underlying AI agents remain critical, under-examined challenges. In this work, we introduce AutoLabs, a self-correcting, multi-agent architecture designed to autonomously translate natural-language instructions into executable protocols for a high-throughput liquid handler. The system engages users in dialogue, decomposes experimental goals into discrete tasks for specialized agents, performs tool-assisted stoichiometric calculations, and iteratively self-corrects its output before generating a hardware-ready file. We present a comprehensive evaluation framework featuring five benchmark experiments of increasing complexity, from simple sample preparation to multi-plate timed syntheses. Through a systematic ablation study of 20 agent configurations, we assess the impact of reasoning capacity, architectural design (single- vs. multi-agent), tool use, and self-correction mechanisms. Our results demonstrate that agent reasoning capacity is the most critical factor for success, reducing quantitative errors in chemical amounts (nRMSE) by over 85% in complex tasks. When combined with a multi-agent architecture and iterative self-correction, AutoLabs approaches expert-authored reference procedures on the benchmark (F1-score > 0.89) on challenging multi-plate syntheses. These findings establish a clear blueprint for developing robust and trustworthy AI partners for autonomous laboratories, highlighting the synergistic effects of modular design, advanced reasoning, and self-correction to ensure both performance and reliability in high-stakes scientific applications. Code: https://github.com/pnnl/autolabs

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗