SEARCH · Engineering Papers

Results for “search engine”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 937 records · Page 52

All sources NASA NTRS DOE OSTI

DOE OSTI2020

Robotic Assisted Non-Destructive Testing (NDT)

Mission Statement: Concrete wall characterization for structural integrity evaluation of H-Canyon exhaust tunnel using a remote controlled robotic arm with a NDT Concrete Instrument. Challenges: Rough and Curved Surfaces, Remote location. The UR5 is a collaborative robotic arm capable of: Payload: 11 lbs (5 kg), Reach: 33.5 in (850 mm), Footprint: 5.87 in (149 mm) diameter, Weight: 45.4 Ibs (20.6 kg). The force torque sensor reacts to a set value inputted by the user that can be utilized for sensitive products. This allows the UR5 to react to surfaces using the built-in function to search for a wall and orient itself normal to that plane. The Proceq Pundit 250 array is a nondestructive device that the user applies against a concrete wall and scans using ultrasonic transducers. The back wall and other defects can be show through the touchscreen. A fixture to hold the Proceq Pundit 250 array onto the UR5 robotic arm was designed and 3D printed. It includes openings for easy access to the buttons from the Pundit array as well as a reinforced structural design. The scan from the Pundit array can show the back-wall, rebars, and other defects to assist in determining the structural integrity. In order to scan, a program was made to search for the wall first. 1. The UR5 robotic arm will go to the desired location away from the wall. 2. It will then search for the wall by moving towards it slowly. 3. After contact, the force applied will slowly but steadily increase to a preset value inputted by the user. 4. The force torque sensor will use those values and communicate with the UR5 by adjusting the arm to become normal to the wall. 5. The arm will then apply force to push back the transducers on the Pundit array to avoid gaps. 6. It will then prompt the user to collect data and wait until finished. The UR5 teach pendant allows the user to program the robot to perform automated and repeatable tasks. It includes a free drive mode which allows the user to move the arm to a desired position by hand. The user can also apply restricted planes which include either stopped or slower motion to ensure safety. The concrete sample being tested includes rebars and different grades of surface roughness to understand the readings from the Proceq Pundit 250 array. Desired concrete samples are currently being fabricated that will simulate the terrain being tested.

3D PRINTING↗

DOE OSTI2021

A systematic comparison of machine learning methods for modeling of dynamic processes applied to combustion emission rate modeling

Ten established, data-driven dynamic algorithms are surveyed and a practical guide for understanding these methods generated. Existing Python programming packages for implementing each algorithm are acknowledged, and the model equations necessary for prediction are presented. A case study on a coal-fired power plant’s NO x emission rates is performed, directly comparing each modeling method’s performance on a mutual system. Each model is evaluated by its root mean squared error (RMSE) on out-of-sample future horizon predictions. Optimal hyperparameters are identified using either an exhaustive search or genetic algorithm. The top five model structures of each method are used to recursively predict future NO x emission rates over a 60-step time horizon. The RMSE at each future timestep is determined, and the recursive output prediction trends compared against measurements in time. The GRU neural network is identified as the best candidate for representing the system, demonstrating accurate and stable predictions across the future horizon by all considered models, while satisfactory performance was observed in several of the ARX/NARX formulations. Finally, these efforts have contributed 1) a concise resource of multiple proven dynamic machine learning methods, 2) a practical guide explaining the use of these methods, effectively lowering the “barrier-to-entry” of deploying such models in control systems, 3) a comparison study evaluating each method’s performance on a mutual system, 4) demonstration of accurate multi-timestep emissions modeling suitable for systems-level control, and 5) generalizable results demonstrating the suitability of each method for prediction over a multi-step future horizon to other complex dynamic systems.

42 ENGINEERING↗

DOE OSTI2023

Experiential findings for sustainable software ecosystems to support experimental and observational science

In the search for a sustainable approach for software ecosystems that supports experimental and observational science (EOS) across Oak Ridge National Laboratory (ORNL), we conducted a survey to understand the current and future landscape of EOS software and data. This paper describes the survey design we used to identify significant areas of interest, gaps, and potential opportunities, followed by a discussion on the obtained responses. The survey formulates questions about project demographics, technical approach, and skills required for the present and the next five years. Here, the study was conducted among 38 ORNL participants between June and July of 2021 and followed the required guidelines for human subjects training. We plan to use the collected information to help guide a vision for sustainable, community-based, and reusable scientific software ecosystems that need to adapt effectively to: (i) the evolving landscape of heterogeneous hardware in the next generation of instruments and computing (e.g. edge, distributed, accelerators), and (ii) data management requirements for data-driven science using artificial intelligence.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS2017

Nano Icy Moons Propellant Harvester

As one of just a few bodies identified in the solar system with a liquid ocean, Europa has become a top priority in the search for life outside of Earth. However, cost estimates for exploring Europa have been prohibitively expensive, with estimates of a NASA Flagship class orbiter and lander approaching $5 billion. ExoTerra's NIMPH offers an affordable solution that can not only land, but return a sample from the surface to Earth. NIMPH combines solar electric propulsion (SEP) technologies being developed for the asteroid redirect mission and microsatellite electronics to reduce the cost of a full sample return mission below $500 million. A key to achieving this order-of-magnitude cost reduction is minimizing the initial mass of the system. The cost of any mission is directly proportional to its mass. By keeping the mission within the constraints of an Atlas V 551 launch vehicle versus an SLS, we can significantly reduce launch costs. To achieve this we reduce the landed mass of the sample return lander, which is the largest multiplier of mission mass, and shrink propellant mass through high efficiency SEP and gravity assists. The NIMPH projects first step in reducing landed mass focuses on development of a micro-In Situ Resource Utilization (micro-ISRU) system. ISRU allows us to minimize landed mass of a sample return mission by converting local ice into propellants. The project reduces the ISRU system to a CubeSat-scale package that weighs just 1.74 kg and consumes just 242 W of power. We estimate that use of this ISRU vs. an identical micro-lander without ISRU reduces fuel mass by 45 kg. As the dry mass of the lander grows for larger missions, these savings scale exponentially. Taking full advantage of the micro-ISRU system requires the development of a micro-liquid oxygen-liquid hydrogen engine. The micro-liquid oxygen-liquid hydrogen engine is tailored for the mission by scaling it to match the scale of the micro-lander and the low gravity of the target moon. We also tailor the engine for a near stoichiometric mixture ratio of 7.5. Most high-performance liquid oxygen-liquid hydrogen engines inject extra liquid hydrogen to lower the average molecular weight of the exhaust, which improves specific impulse. However, this extra liquid hydroden requires additional power and processing time on the surface for the ISRU to create. This increases mission cost, and on missions within high radiation environments such as Europa, increases radiation shielding mass. The resulting engine weighs just 1.36 kg and produces 71.5 newton of thrust at 364 s specific impulse. Finally, the mission reduces landed mass by taking advantage of the SEP modules solar power to beam energy to the surface using a collimated laser. This allows us to replace an 45 kg MMRTG with a 2.5 kg resonant array. By using the combination of ISRU, a liquid oxygen-liquid hydrogen engine, and beamed power, we reduce the initial mass of the lander to just 51.5 kg. When combined with an SEP module to ferry the lander to Europa the initial mission mass is just 6397 kg - low enough to be placed on an Earth escape trajectory using an Atlas V 551 launch vehicle. By comparison, we estimate a duplicate lander using an MMRTG and semi-storable propellants such as liquid oxygen-methane would result in an order of magnitude increase in initial lander mass to 445 kg. Attempting to perform the trajectory with a 450 s liquid oxygen-liquid hydrogen engine would increase initial mass to approximately 135,000 kg. Using an Atlas V 1 U.S. Dollar per kg rate to Earth escape value of $27.7k per kg, just the launch savings are over $3.5 billion.

liquid↗

NASA NTRS1999

Interplanetary Microlaser Transponders

The feasibility of an asynchronous (i.e. independently firing) interplanetary laser transponder, capable of ranging between Earth and Mars and using the automated SLR2000 Satellite Laser Ranging (SLR) system as an Earth base station, has been suggested. Since that time, we have received a small amount of discretionary funding to further explore the transponder concept and to develop and test an engineering breadboard. Candidate operational scenarios for acquiring and tracking the opposite laser terminal over interplanetary distances have been developed, and breadboard engineering parameters were chosen to reflect the requirements of an Earth-Mars link Laboratory tests have been devised to simulate the Earth- Mars link between two independent SLR2000 transceivers and to demonstrate the transfer of range and time in single photon mode. The present paper reviews the concept of the asynchronous microlaser transponder, the transponder breadboard design, an operational scenario recently developed for an asteroid rendezvous, and the laboratory test setup. The optical head of the transponder breadboard fits within a cylinder roughly 15 cm in diameter and 32 cm in length and is mounted in a commercial two axis gimbal driven by two computer-controlled stepper motors which allows the receiver optical axis to be centered on a simulated Earth image. The optical head is built around a small optical bench which supports a 14.7 cm diameter refractive telescope, a prototype 2 kHz SLR2000 microlaser transmitter, a quadrant microchannel plate photomultiplier (MCP/PMT), a CCD array camera, spatial and spectral filters, assorted lenses and mirrors, and protective covers and sun shields. The microlaser is end-pumped by a fiber-coupled diode laser array. An annular mirror is employed as a passive transmit/receive (T/R) switch in an aperture-sharing arrangement wherein the transmitted beam passes through the central hole and illuminates only the central 2.5 cm of the common telescope (adequate to achieve a 10 arcsecond full laser beam divergence) while the receiver uses the remainder of the 14.7 cm aperture. Additional electronic instrumentation includes the diode pump array and associated heat sink and current drivers, rubidium frequency standard, timing distribution module, range gate generator, a recently developed all-digital correlation range receiver, and system computer. Acquisition of the opposite transponder terminal requires a search within a three-dimensional volume determined by the initial pointing uncertainty and a maximum 500 microsecond uncertainty in the laser time of fire at the opposite terminal for totally uncorrelated Earth and spacecraft clocks. The angular search is aided by a sensitive CCD array capable of imaging the Earth, Moon, and surrounding stars within the nominal + 0.5 degree cone of uncertainty associated with the initial pointing of a spacecraft body or microwave communications dish.

Degnan, John J.↗

DOE OSTI2022

Physics makes the difference: Bayesian optimization and active learning via augmented Gaussian process

Abstract Both experimental and computational methods for the exploration of structure, functionality, and properties of materials often necessitate the search across broad parameter spaces to discover optimal experimental conditions and regions of interest in the image space or parameter space of computational models. The direct grid search of the parameter space tends to be extremely time-consuming, leading to the development of strategies balancing exploration of unknown parameter spaces and exploitation towards required performance metrics. However, classical Bayesian optimization (BO) strategies based on the Gaussian process (GP) do not readily allow for the incorporation of the known physical behaviors or past knowledge. Here we explore a hybrid optimization/exploration algorithm created by augmenting the standard GP with a structured probabilistic model of the expected system’s behavior. This approach balances the flexibility of the non-parametric GP approach with a rigid structure of physical knowledge encoded into the parametric model. The fully Bayesian treatment of the latter allows additional control over the optimization via the selection of priors for the model parameters. The method is demonstrated for a noisy version of a standard univariate test function used to evaluate optimization algorithms and further extended to physical lattice models. This methodology is expected to be universally suitable for injecting prior knowledge in the form of physical models and past data in the BO framework.

42 ENGINEERING↗

DOE OSTI2025

Measurement of the liquid argon scintillation pulse shape using differentiable simulation in the coherent CAPTAIN-Mills experiment

The coherent CAPTAIN-Mills (CCM) experiment is a liquid argon (LAr) light collection detector searching for MeV-scale neutrino and beyond Standard Model physics signatures. Two hundred eight-inch photomultiplier tubes instrument the seven-ton fiducial volume with 50% photocathode coverage to detect light produced by charged particles. CCM’s light-based approach reduces requirements of LAr purity, compared to other detection technologies, such that sub-MeV particles can be reliably detected without additional LAr filtration and with 𝒪⁡(1) parts per million of common contaminants. We present a measurement of LAr light production and propagation parameters, with uncertainties, obtained from a sample of MeV-scale electromagnetic events. The optimization of this high-dimensional parameter space was facilitated by a differentiable optical photon Monte Carlo simulation and detailed photomultiplier tube response characterization. This result accurately predicts the timing and spatial distribution of light due to scintillation and Cherenkov emission in the detector. This is the first description of photon propagation in LAr to include several effects, including anomalous dispersion of the index of refraction near the ultraviolet resonance, Mie scattering from impurities, and Cherenkov light production.

Cherenkov detectors↗

DOE OSTI2024

Integrated ab initio modelling of atomic ordering and magnetic anisotropy for design of FeNi-based magnets

We describe an integrated modelling approach to accelerate the search for novel, single-phase, multicomponent materials with high magnetocrystalline anisotropy (MCA). For a given system we predict the nature of atomic ordering, its dependence on the magnetic state, and then proceed to describe the consequent MCA, magnetisation, and magnetic critical temperature (Curie temperature). Crucially, within our modelling framework, the same ab initio description of a material’s electronic structure determines all aspects. We demonstrate this holistic method by studying the effects of alloying additions in FeNi, examining systems with the general stoichiometries Fe 4 Ni 3 X and Fe 3 Ni 4 X, for additives including X = Pt, Pd, Al, and Co. The atomic ordering behaviour predicted on adding these elements, fundamental for determining a material’s MCA, is rich and varied. Equiatomic FeNi has been reported to require ferromagnetic order to establish the tetragonal L1 0 order suited for significant MCA. Our results show that when alloying additions are included in this material, annealing in an applied magnetic field and/or below a material’s Curie temperature may also promote tetragonal order, along with an appreciable effect on the predicted hard magnetic properties.

36 MATERIALS SCIENCE↗

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 OSTI2024

Data for "Integrated ab initio modelling of atomic order and magnetic anisotropy for rare-earth-free magnet design: effects of alloying additions in L1 0 FeNi."

36 MATERIALS SCIENCE↗

DOE OSTI2025

Searching for a Pulse: Evaluating the Use of Rapid DC Pulses for Diagnosing Battery Health, State-of-Charge, and Safety

Rapid electrochemical diagnostics, like DC pulse sequences or electrochemical impedance spectroscopy, are known to be useful for capacity prediction. However, it is unclear how previous results will map to different cell types and use cases and whether rapid diagnostics are useful for remaining useful life prediction or for detecting potential safety issues. To that end, we have collected a data set with ∼50,000 DC pulse measurements from four types of commercial lithium-ion batteries to enable training of state-of-charge, health, and safety diagnostic models via machine-learning. We demonstrate that 120-second DC pulse sequences can be used to predict capacity with 2%–9% average error, which can separate high- from low-capacity cells with only a 0.3% false positive rate but is not accurate enough to estimate remaining useful life. We also find that no safety related targets can be accurately predicted, highlighting the critical need for other non-invasive methods to diagnose battery safety.

25 ENERGY STORAGE↗

DOE OSTI2021

Exploring the elastic and electronic properties of chromium molybdenum diboride alloys

We perform first-principles calculations to study the structural, mechanical, thermal, electronic, and magnetic properties of Cr 1-x Mo x B 2 for x = 0.25, 0.33, 0.50, 0.67 and 0.75. Based on structural search methods, we determine the ground-state structure for each concentration. The ternaries are either monoclinic (x = 0.25, 0.75) or trigonal (x = 0.33, 0.50, 0.67). The calculated mechanical properties reveal that the strength of Cr 1-x Mo x B 2 is maximized for x = 0.50. Cr 0.5 Mo 0.5 B 2 exhibits excellent mechanical properties (B = 298 GPa, Y = 558 GPa, G = 235 Gpa, ν = 0.19, H v =27 GPa), surpassing those of β-MoB 2 at a lower cost. All of these ternaries are hard alloys with Vickers hardness greater than 24 GPa. Chemical bonding analysis demonstrates that the strength of the new compounds is related to the alternating planar and buckled B-B layers, as well as the strong TM-B bonds. Finally, the enhanced strength of Cr 0.5 Mo 0.5 B 2 is a consequence of the high density of strong interlayer Cr-Mo metallic bonds around the Fermi level.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Nonresonant Raman Control of Ferroelectric Polarization

Important advances is recently made in the search for materials with complex multi-phase landscapes that host photoinduced metastable collective states with exotic functionalities. In almost all cases so far, the desired phases are accessed by exploiting light–matter interactions via the imaginary part of the dielectric function through above-bandgap or resonant mode excitation. Nonresonant Raman excitation of coherent modes is experimentally observed and proposed for dynamic material control, but the resulting atomic excursion is limited to perturbative levels. Here, in this study, this challenge is overcome by employing nonresonant ultrashort pulses with low photon energies well below the bandgap. Using mid-infrared pulses, ferroelectric reversal is induced in lithium niobate, and the large-amplitude mode displacements are characterized through femtosecond stimulated Raman scattering and second harmonic generation. This approach, validated by first-principle calculations, defines a novel method for synthesizing hidden phases with unique functional properties and manipulating complex energy landscapes at reduced energy consumption and ultrafast speeds.

Shi, Jiaojian [Stanford Univ., CA (United States);↗

DOE OSTI2021

Fast Local Spatial Verification for Feature-Agnostic Large-Scale Image Retrieval

Images from social media can reflect diverse viewpoints, heated arguments, and expressions of creativity, adding new complexity to retrieval tasks. Researchers working on Content-Based Image Retrieval (CBIR) have traditionally tuned their algorithms to match filtered results with user search intent. However, we are now bombarded with composite images of unknown origin, authenticity, and even meaning. With such uncertainty, users may not have an initial idea of what the search query results should look like. For instance, hidden people, spliced objects, and subtly altered scenes can be difficult for a user to detect initially in a meme image, but may contribute significantly to its composition. It is pertinent to design systems that retrieve images with these nuanced relationships in addition to providing more traditional results, such as duplicates and near-duplicates — and to do so with enough efficiency at large scale. In this work, we propose a new approach for spatial verification that aims at modeling object-level regions using image keypoints retrieved from an image index, which is then used to accurately weight small contributing objects within the results, without the need for costly object detection steps. We call this method the Objects in Scene to Objects in Scene (OS2OS) score, and it is optimized for fast matrix operations, which can run quickly on either CPUs or GPUs. It performs comparably to state-of-the-art methods on classic CBIR problems (Oxford 5K, Paris 6K, and Google-Landmarks), and outperforms them in emerging retrieval tasks such as image composite matching in the NIST MFC2018 dataset and meme-style imagery from Reddit.

42 ENGINEERING↗

NASA NTRS2011

Shooting the Moon

This story is about an unlikely NASA mission to the Moon. It was unlikely because it was started with far too little time and too-little money to complete. It was unlikely because it was able to take chances to accept risk of failure. It was unlikely because it was searching for the unthinkable: water-ice on the moon... Figure 1-1: LCROSS Mission. The mission of the Lunar CRater Observation and Sensing Satellite (LCROSS) was to investigate the possibility of water ice in craters on the Moon s poles. This is certainly an interesting scientific topic in itself, but I intend to focus on the compelling experience of managing the LCROSS Project in the context of this storied Agency. Perhaps most interesting are the implications this story has for managing any development effort, lunar or not, and working a balance to achieve success. NASA is by design a risk-taking agency within the US Government. It could be argued that NASA s purpose in the aerospace community is to take on the really big challenges that either the corporate world can t afford, are not yet profitable endeavors, or are just too risky for private corporations to entertain. However, expectations of the Agency have evolved. A combination of grim human tragedies and some very public cost and schedule overruns have challenged the public s and Congress s tolerance for risk-taking within the Agency. NASA, which is supposed to be in the business of taking risks to do bold, difficult things, has become less and less able to do so within its cost framework. Yet effectively replacing prudent risk management with attempts to "risk-eliminate" is completely unaffordable. So where does risk-taking fit within the Agency, or within private/corporate organizations for that matter? Where astronauts play there is clearly concern about risk. When an organization puts humans in harm s way, it is understandably going to take extra effort to assure nobody gets hurt. Doing so, of course, costs money - a lot of money to pay for labor and hardware which is attempting to assure nothing will go wrong. Sophisticated designs, with doubly- or triply-redundant systems, extensive testing to verify those systems, and numerous engineering test units built to learn and evolve a hardware design, all drive the cost and time required to implement. Human spaceflight is an expensive business because of the exceptional system complexity and levels of assurance required for human space travel. What about missions that do not involve human spaceflight? What about missions whose potential failure will not take a human life, whose costs are small and whose urgency and importance are limited by design? A portfolio consisting of this type of mission can be designed to be risk tolerant, not requiring large expenditures to guarantee against failure. With the money saved, the number of missions that can be executed within the portfolio grows, or the total cost of the portfolio can be reduced. The NASA LCROSS mission is a pathfinder example of a low-cost, quick turn-around mission which struck a balance on mission risk, while accomplishing big objectives, like defining how we understand the Moon.

Andrews, Daniel R.↗

DOE OSTI2021

A Simple Criterion for Feasibility of Heat Integration between Distillation Streams Based on Relative Volatilities

In a multicomponent distillation configuration, there are numerous sources and sinks of heat, and a potential way to reduce the heat duty requirement is to perform heat integration. Unfortunately, an algorithmic search of the optimal heat integration opportunities is intractable when the required temperatures of intermediate streams are computed via complex models. Instead, in this work, we introduce pressure-scaled pseudo relative volatility, a new metric to compare stream temperatures. Here, we justify the use of pseudo relative volatility by proving that this variable is a monotonically increasing function of the liquid fraction in a saturated mixture stream. Using this metric, we derive a shortcut criterion to check the feasibility of various heat integration opportunities, such as thermal coupling via heat transfer (TCH). The advantage of this approach is that it circumvents the need for explicit temperatures and instead relies on composition, component relative volatilities, and pressure—quantities that are readily available in shortcut models for optimization of distillation configurations. Leveraging this fact, we propose a new optimization framework that identifies feasible TCHs that we consider within the formulation while minimizing the total heat duty of a distillation configuration. We demonstrate, on a few examples, that our formulation can identify heat duty efficient configurations, some of which are multieffect configurations. Using this methodology, we discover configurations that are not only simpler than the fully thermally coupled (FTC) configuration but also have a much lower heat duty.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2023

Selecting Minimal Motion Primitive Libraries with Genetic Algorithms

Motion primitives allow for application of discrete search algorithms to rapidly produce trajectories in complex continuous space. The maneuver automaton (MA) provides an elegant formulation for creating a primitive library based on trims and maneuvers. However, performance is fundamentally limited by the contents of the primitive library. If the library is too sparse, performance can be poor in terms of path cost, whereas a library that is too large can increase run time. This work outlines new methods for using genetic algorithms to prune a primitive library. The proposed methods balance the path cost and planning time while maintaining the reachability of the MA. The genetic algorithm in this paper evaluates and mutates populations of motion primitive libraries to optimize both objectives. Here, we illustrate the performance of these methods with a simulated study using a nonlinear medium-fidelity F-16 model. We optimize a library with the presented algorithm for obstacle-free navigation and a nap-of-the-Earth navigation task. In the obstacle-free navigation task, we show a tradeoff of a 10.16% higher planning cost for a 96.63% improvement in run time. In the nap-of-the-Earth task, we show a tradeoff of a 9.712% higher planning cost for a 92.06% improvement in run time.

42 ENGINEERING↗

NASA NTRS2007

Automated Knowledge Discovery From Simulators

A computational method, SimLearn, has been devised to facilitate efficient knowledge discovery from simulators. Simulators are complex computer programs used in science and engineering to model diverse phenomena such as fluid flow, gravitational interactions, coupled mechanical systems, and nuclear, chemical, and biological processes. SimLearn uses active-learning techniques to efficiently address the "landscape characterization problem." In particular, SimLearn tries to determine which regions in "input space" lead to a given output from the simulator, where "input space" refers to an abstraction of all the variables going into the simulator, e.g., initial conditions, parameters, and interaction equations. Landscape characterization can be viewed as an attempt to invert the forward mapping of the simulator and recover the inputs that produce a particular output. Given that a single simulation run can take days or weeks to complete even on a large computing cluster, SimLearn attempts to reduce costs by reducing the number of simulations needed to effect discoveries. Unlike conventional data-mining methods that are applied to static predefined datasets, SimLearn involves an iterative process in which a most informative dataset is constructed dynamically by using the simulator as an oracle. On each iteration, the algorithm models the knowledge it has gained through previous simulation trials and then chooses which simulation trials to run next. Running these trials through the simulator produces new data in the form of input-output pairs. The overall process is embodied in an algorithm that combines support vector machines (SVMs) with active learning. SVMs use learning from examples (the examples are the input-output pairs generated by running the simulator) and a principle called maximum margin to derive predictors that generalize well to new inputs. In SimLearn, the SVM plays the role of modeling the knowledge that has been gained through previous simulation trials. Active learning is used to determine which new input points would be most informative if their output were known. The selected input points are run through the simulator to generate new information that can be used to refine the SVM. The process is then repeated. SimLearn carefully balances exploration (semi-randomly searching around the input space) versus exploitation (using the current state of knowledge to conduct a tightly focused search). During each iteration, SimLearn uses not one, but an ensemble of SVMs. Each SVM in the ensemble is characterized by different hyper-parameters that control various aspects of the learned predictor - for example, whether the predictor is constrained to be very smooth (nearby points in input space lead to similar output predictions) or whether the predictor is allowed to be "bumpy." The various SVMs will have different preferences about which input points they would like to run through the simulator next. SimLearn includes a formal mechanism for balancing the ensemble SVM preferences so that a single choice can be made for the next set of trials.

Burl, Michael↗