SEARCH · Search NASA

Results for “chain architecture”

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 91 records · Page 5

All sources NASA NTRS DOE OSTI

NASA NTRS2007

Miniature L-Band Radar Transceiver

A miniature L-band transceiver that operates at a carrier frequency of 1.25 GHz has been developed as part of a generic radar electronics module (REM) that would constitute one unit in an array of many identical units in a very-large-aperture phased-array antenna. NASA and the Department of Defense are considering the deployment of such antennas in outer space; the underlying principles of operation, and some of those of design, also are applicable on Earth. The large dimensions of the antennas make it advantageous to distribute radio-frequency electronic circuitry into elements of the arrays. The design of the REM is intended to implement the distribution. The design also reflects a requirement to minimize the size and weight of the circuitry in order to minimize the weight of any such antenna. Other requirements include making the transceiver robust and radiation-hard and minimizing power demand. Figure 1 depicts the functional blocks of the REM, including the L-band transceiver. The key functions of the REM include signal generation, frequency translation, amplification, detection, handling of data, and radar control and timing. An arbitrary-waveform generator that includes logic circuitry and a digital-to-analog converter (DAC) generates a linear-frequency-modulation chirp waveform. A frequency synthesizer produces local-oscillator signals used for frequency conversion and clock signals for the arbitrary-waveform generator, for a digitizer [that is, an analog-to-digital converter (ADC)], and for a control and timing unit. Digital functions include command, timing, telemetry, filtering, and high-rate framing and serialization of data for a high-speed scientific-data interface. The aforementioned digital implementation of filtering is a key feature of the REM architecture. Digital filters, in contradistinction to analog ones, provide consistent and temperature-independent performance, which is particularly important when REMs are distributed throughout a large array. Digital filtering also enables selection among multiple filter parameters as required for different radar operating modes. After digital filtering, data are decimated appropriately in order to minimize the data rate out of an antenna panel. The L-band transceiver (see Figure 2) includes a radio-frequency (RF)-to-baseband down-converter chain and an intermediate- frequency (IF)-to-RF up-converter chain. Transmit/receive (T/R) switches enable the use of a single feed to the antenna for both transmission and reception. The T/R switches also afford a built-in test capability by enabling injection of a calibration signal into the receiver chain. In order of decreasing priority, components of the transceiver were selected according to requirements of radiation hardness, then compactness, then low power. All of the RF components are radiation-hard. The noise figure (NF) was optimized to the extent that (1) a low-noise amplifier (LNA) (characterized by NF < 2 dB) was selected but (2) the receiver front-end T/R switches were selected for a high degree of isolation and acceptably low loss, regardless of the requirement to minimize noise.

McWatters, Dalia↗

NASA NTRS2008

Ensemble: an Architecture for Mission-Operations Software

Ensemble is the name of an open architecture for, and a methodology for the development of, spacecraft mission operations software. Ensemble is also potentially applicable to the development of non-spacecraft mission-operations- type software. Ensemble capitalizes on the strengths of the open-source Eclipse software and its architecture to address several issues that have arisen repeatedly in the development of mission-operations software: Heretofore, mission-operations application programs have been developed in disparate programming environments and integrated during the final stages of development of missions. The programs have been poorly integrated, and it has been costly to develop, test, and deploy them. Users of each program have been forced to interact with several different graphical user interfaces (GUIs). Also, the strategy typically used in integrating the programs has yielded serial chains of operational software tools of such a nature that during use of a given tool, it has not been possible to gain access to the capabilities afforded by other tools. In contrast, the Ensemble approach offers a low-risk path towards tighter integration of mission-operations software tools.

Norris, Jeffrey↗

NASA NTRS2006

NASA Space Exploration Logistics Workshop Proceedings

As NASA has embarked on a new Vision for Space Exploration, there is new energy and focus around the area of manned space exploration. These activities encompass the design of new vehicles such as the Crew Exploration Vehicle (CEV) and Crew Launch Vehicle (CLV) and the identification of commercial opportunities for space transportation services, as well as continued operations of the Space Shuttle and the International Space Station. Reaching the Moon and eventually Mars with a mix of both robotic and human explorers for short term missions is a formidable challenge in itself. How to achieve this in a safe, efficient and long-term sustainable way is yet another question. The challenge is not only one of vehicle design, launch, and operations but also one of space logistics. Oftentimes, logistical issues are not given enough consideration upfront, in relation to the large share of operating budgets they consume. In this context, a group of 54 experts in space logistics met for a two-day workshop to discuss the following key questions: 1. What is the current state-of the art in space logistics, in terms of architectures, concepts, technologies as well as enabling processes? 2. What are the main challenges for space logistics for future human exploration of the Moon and Mars, at the intersection of engineering and space operations? 3. What lessons can be drawn from past successes and failures in human space flight logistics? 4. What lessons and connections do we see from terrestrial analogies as well as activities in other areas, such as U.S. military logistics? 5. What key advances are required to enable long-term success in the context of a future interplanetary supply chain? These proceedings summarize the outcomes of the workshop, reference particular presentations, panels and breakout sessions, and record specific observations that should help guide future efforts.

deWeek, Oliver↗

NASA NTRS2013

Digitally Calibrated TR Modules Enabling Real-Time Beamforming SweepSAR Architectures

SweepSAR, a novel radar architecture that depends on a DBF (digital beamforming) array, requires calibration accuracies that are order(s) of magnitude greater than is possible with traditional techniques, such as a priori characterization of TR (transmit/receive) modules in thermal vacuum chambers, or simple loop-back of the calibration signal. The advantages of a SweepSAR architecture are so great that it is worth applying significant resources to calibration efforts. Due to the nature of the DBF, each channel contains a digitizer and very powerful digital processor. Each channel can independently digitize (with the digitizer) and analyze (with the processor) its channel's unique calibration signal, and extract the relevant calibration parameters, namely channel gain and channel phase delay commonly referred to as the gain (or amplitude) and phase of the channel. Using the processor, each channel's gain and phase can theoretically be estimated with arbitrary precision through averaging a sufficiently large number of samples. Systematic errors and the changing gain and phase of the channels, typically due to temperature drifts, limits how long the averaging can occur, which limits the precision of the calibration estimate. However, results indicate that calibration knowledge of both the transmit and receive chains of each TR module can be improved by one or two orders of magnitude. Due to the digital nature of the receiver data, the channel's gain and phase may be corrected by a similar amount, while the transmit chain can only be corrected in a traditional manner. To implement Sweep SAR, the order of magnitude improvement in the knowledge of the channel's gain and phase is needed, and the control of the receiver to a similar level is required. Inherent to the DBF array is the individual digitization of each of the array's receiver channels. Current systems typically combine all of the analog signals in the array into one or two analog channels, which are then digitized and processed. All signal conditioning performed prior to digitization is done using analog hardware (which is less precise than digital signal conditioning and dependent on temperature). The DBF digitizes every signal prior to combining, and can therefore analyze and correct received signals, as well as analyze signals that are being transmitted through analog hardware (by sampling a copy and digitizing). Each channel of a DBF also has a powerful processor. With this combination, one is able to digitize, analyze, and correct each channel prior to its being combined. A unique factor is the ability to digitize and analyze (in real time) each of the array's channels independently, allowing one to achieve unprecedented knowledge of each channel's performance (gain and phase), and since the combining is done digitally, each receive channel can be corrected prior to combining. This enables an unprecedented level of accuracy and control through onboard processing. SweepSAR promises significant increases in instrument capability for solid earth and biomass remote sensing, while reducing mission mass and cost. This new instrument concept requires new methods for calibrating the multiple channels, which must be combined onboard, in real time. New methods are being developed for digitally calibrating digital beam-forming arrays to reduce development time, risk, and cost of precision calibrated TR modules for array architectures by accurately tracking modules' characteristics through closed-loop digital calibration, thus tracking systematic changes regardless of temperature.-

Hoffman, James P.↗

NASA NTRS1985

A study of topologies and protocols for fiber optic local area network

The emergence of new applications requiring high data traffic necessitates the development of high speed local area networks. Optical fiber is selected as the transmission medium due to its inherent advantages over other possible media and the dual optical bus architecture is shown to be the most suitable topology. Asynchronous access protocols, including token, random, hybrid random/token, and virtual token schemes, are developed and analyzed. Exact expressions for insertion delay and utilization at light and heavy load are derived, and intermediate load behavior is investigated by simulation. A new tokenless adaptive scheme whose control depends only on the detection of activity on the channel is shown to outperform round-robin schemes under uneven loads and multipacket traffic and to perform optimally at light load. An approximate solution to the queueing delay for an oscillating polling scheme under chaining is obtained and results are compared with simulation. Solutions to the problem of building systems with a large number of stations are presented, including maximization of the number of optical couplers, and the use of passive star/bus topologies, bridges and gateways.

Yeh, C.↗

NASA NTRS2022

A Modular Framework for Integrating and Visualizing Telemetry for Mars 2020 Rover Mechanism Operations

The analysis of mechanism telemetry requires a wide variety of tools to quickly and effectively assess spacecraft state, capture long-term trends in system performance, and identify and track anomalous events. Such analysis often requires spacecraft telemetry to first be transformed into derived fields and aggregated statistics before operators can begin their analysis. In past missions, aspects of this process have been automated, but operators were expected to use their own tools and procedures to understand and visualize the data, which led to redundant and inconsistent tools and processes. The Mech Data Tools Python library (MDT) was developed to provide a flexible, unified tool set for operators to extract and analyze mechanism telemetry over the life of the Mars 2020 surface mission. MDT consists of a set of configurable components that implement standard interfaces for ingesting input and producing output. Components can be chained together to form a data processing pipeline. Data are ingested from several sources within the greater Mars 2020 cloud infrastructure and stored in pandas DataFrames, which allows users to leverage the data manipulation capabilities present within the widely-used pandas library. Visualization capabilities are provided through the Plotly library, which generates interactive plots for users to interpret. Following the beginning of Mars 2020 surface operations, usage of MDT has spread to all mechanism-focused subsystems and has demonstrated great utility in analyzing early surface activities. This paper describes MDT’s evolution from heritage mechanism telemetry tools, the critical architecture decisions and challenges faced over MDT’s two years of development, and current applications of MDT in support of mechanism operations.

Wolsieffer, Ben↗

NASA NTRSDate not supplied

Ballistic Lunar Return Trajectories for Sustainable Cargo Return and Entry System Technology Development

As part of the sustained approach for the lunar Gateway and Artemis program as a whole, NASA is extending the logistics supply chain beyond low Earth orbit and to the Moon. This supply chain includes the possibility of lunar sample and cargo return. To enable these future return missions, the possibility for incorporating demonstration payloads including various entry, descent, and landing technologies is examined. Possible implementations include deployable entry vehicles, high speed sample return capsules, aeroassist technologies, and control technologies for guided hypersonic flight. The mission concepts utilize the secondary payload capabilities provided on a relatively low-cost logistic module. The logistic module may perform close Earth flyby, pointing, and release of return systems prior to disposal, with reentry velocities for payloads on the order of 11.5 km/sec. In this paper, we study the various return mission architectures available during the Artemis program to define the range of mission possibilities. Potential options include a reusable logistics module using a hypersonic inflatable aerodynamic decelerator, externally mounted entry system vehicle technologies, and a reusable sample return vehicle using a similar inflatable architecture with a feathered configuration with applied aerodynamic control. In these cases, the use of a ballistic lunar return trajectory is assumed, and a sensitivity analysis of midcourse corrections and the possibility of a lunar gravity assist for refining Earth entry interface points is provided. Furthermore, a novel controller for controlling a vehicle during reentry imposing heating limitations is introduced.

Matthew M. Wittal↗

NASA NTRS1994

Cascaded clocks measurement and simulation findings

This paper will examine aspects related to network synchronization distribution and the cascading of timing elements. Methods of timing distribution have become a much debated topic in standards forums and among network service providers (both domestically and internationally). Essentially these concerns focus on the need to migrate their existing network synchronization plans (and capabilities) to those required for the next generation of transport technologies (namely, the Synchronous Digital Hierarchy (SDH), Synchronous Optical Networks (SONET), and Asynchronous Transfer Mode (ATM). The particular choices for synchronization distribution network architectures are now being evaluated and are demonstrating that they can indeed have a profound effect on the overall service performance levels that will be delivered to the customer. The salient aspects of these concerns reduce to the following: (1) identifying that the devil is in the details of the timing element specifications and the distribution of timing information (i.e., small design choices can have a large performance impact); (2) developing a standardized method of performance verification that will yield unambiguous results; and (3) presentation of those results. Specifically, this will be done for two general cases: an ideal input, and a noisy input to a cascaded chain of slave clocks.

Chislow, Don↗

NASA NTRS1993

Parallel O(log n) algorithms for open- and closed-chain rigid multibody systems based on a new mass matrix factorization technique

In this paper, parallel O(log n) algorithms for computation of rigid multibody dynamics are developed. These parallel algorithms are derived by parallelization of new O(n) algorithms for the problem. The underlying feature of these O(n) algorithms is a drastically different strategy for decomposition of interbody force which leads to a new factorization of the mass matrix (M). Specifically, it is shown that a factorization of the inverse of the mass matrix in the form of the Schur Complement is derived as M(exp -1) = C - B(exp *)A(exp -1)B, wherein matrices C, A, and B are block tridiagonal matrices. The new O(n) algorithm is then derived as a recursive implementation of this factorization of M(exp -1). For the closed-chain systems, similar factorizations and O(n) algorithms for computation of Operational Space Mass Matrix lambda and its inverse lambda(exp -1) are also derived. It is shown that these O(n) algorithms are strictly parallel, that is, they are less efficient than other algorithms for serial computation of the problem. But, to our knowledge, they are the only known algorithms that can be parallelized and that lead to both time- and processor-optimal parallel algorithms for the problem, i.e., parallel O(log n) algorithms with O(n) processors. The developed parallel algorithms, in addition to their theoretical significance, are also practical from an implementation point of view due to their simple architectural requirements.

Fijany, Amir↗

NASA NTRS1999

Genetic ablation of root cap cells in Arabidopsis

The root cap is increasingly appreciated as a complex and dynamic plant organ. Root caps sense and transmit environmental signals, synthesize and secrete small molecules and macromolecules, and in some species shed metabolically active cells. However, it is not known whether root caps are essential for normal shoot and root development. We report the identification of a root cap-specific promoter and describe its use to genetically ablate root caps by directing root cap-specific expression of a diphtheria toxin A-chain gene. Transgenic toxin-expressing plants are viable and have normal aerial parts but agravitropic roots, implying loss of root cap function. Several cell layers are missing from the transgenic root caps, and the remaining cells are abnormal. Although the radial organization of the roots is normal in toxin-expressing plants, the root tips have fewer cytoplasmically dense cells than do wild-type root tips, suggesting that root meristematic activity is lower in transgenic than in wild-type plants. The roots of transgenic plants have more lateral roots and these are, in turn, more highly branched than those of wild-type plants. Thus, root cap ablation alters root architecture both by inhibiting root meristematic activity and by stimulating lateral root initiation. These observations imply that the root caps contain essential components of the signaling system that determines root architecture.

Non-NASA Center↗

NASA NTRS2023

Progressively Enabling Earth Independent Medical Operations (EIMO)

This panel presents the findings from a series of Technical Interchange Meetings (TIMs) hosted by the Exploration Medical Capability Element (ExMC) in NASA’s Human Research Program. The topics for the TIMs were derived from a 2-day conference of senior leaders and subject matters experts that collectively outlined a multi-faceted strategy designed to optimize crew health and performance through an increasingly autonomous medical approach. The first abstract in this panel outlines the scope of issues related to data collection, usage, transmission and computing capacity to facilitate EIMO. The second presentation provides an overview of the challenges in developing curricula and advanced training tools to baseline knowledge, skills and abilities (KSA), verify clinical competency and assure retention during prolonged durations inherent in exploration-class missions. An overview of the complicated medical supply and resource chain necessary to facilitate EIMO is provided in the third presentation of this panel. The final presentation in this EIMO panel surveys the breadth and depth of demands on cognitive load expected to be experienced by crew on an exploration mission and proposes strategies to mitigate the prospect of cognitive overload through methods to shift task load from the crew to multi-modal artificial intelligence based medical support systems. Taken together, these presentations summarize the challenges to be expected and potential solution spaces to be explored and developed to progressively enable increasing autonomous medical operations to support crewed missions beyond low earth orbit. Through EIMO focused pre-mission planning, integrated data architecture design, innovative training development and AI-assisted task load management, the gradual transition of medical care and decision making from terrestrial to space-based assets enabling support of astronaut health and performance and reducing overall mission risk is achievable.

Jay Lemery↗

NASA NTRSDate not supplied

Progressively Enabling Earth Independent Medical Operations (EIMO)

John Lemery↗

NASA NTRSDate not supplied

From Research to Reality -- Challenges and Opportunities in Complex System Design

The scope and scale of our interconnected society requires that we view the world through a complex system lens, where numerous parts interact, and emergent behaviors are the norm. Understanding complex systems is critical for decision-making and policy development in domains such as ecological systems, financial markets, supply chains, and global transportation systems, where decision-makers need reliable information to predict the impact of decisions that may play out over decades. Traditional approaches to the research that produces this information are often insufficient, where hypotheses are tested in an isolated environment, and where the results may not carry over to the integrated system. There are a multitude of advancements in system engineering, artificial intelligence, and test and evaluation that are emerging to meet the challenge. Approaches such as agile development, model-based system engineering, design of experiments, large language models, and formal ontologies are providing ways to manage complexity and increase our collective ability to make the changes that we want to see in the world. In this talk I will provide a few examples related to the architecture of the National Airspace System (NAS), where we have investigated the use of large language models, basic formal ontology, and applied category theory to help researchers and system engineers be more effective in this complex design space. Bio: Dr. Ian Levitt’s current research focus is on the complex evolution of the National Airspace System. Prior to joining NASA in 2020, he was with the FAA leading international standards and national laboratory development for the agency. Dr. Levitt earned his PhD in mathematics from Rutgers University in 2009. His mission is to promote a healthy and continuous transformation of society through open information and cooperation.

Ian Levitt↗

NASA NTRS1990

Developing an Intelligent Computer-Aided Trainer

The Payload-assist module Deploys/Intelligent Computer-Aided Training (PD/ICAT) system was developed as a prototype for intelligent tutoring systems with the intention of seeing PD/ICAT evolve and produce a general ICAT architecture and development environment that can be adapted by a wide variety of training tasks. The proposed architecture is composed of a user interface, a domain expert, a training session manager, a trainee model and a training scenario generator. The PD/ICAT prototype was developed in the LISP environment. Although it has been well received by its peers and users, it could not be delivered toe its end users for practical use because of specific hardware and software constraints. To facilitate delivery of PD/ICAT to its users and to prepare for a more widely accepted development and delivery environment for future ICAT applications, we have ported this training system to a UNIX workstation and adopted use of a conventional language, C, and a C-based rule-based language, CLIPS. A rapid conversion of the PD/ICAT expert system to CLIPS was possible because the knowledge was basically represented as a forward chaining rule base. The resulting CLIPS rule base has been tested successfully in other ICATs as well. Therefore, the porting effort has proven to be a positive step toward our ultimate goal of building a general purpose ICAT development environment.

Hua, Grace↗

NASA NTRS1994

Implementation of a fully-balanced periodic tridiagonal solver on a parallel distributed memory architecture

While parallel computers offer significant computational performance, it is generally necessary to evaluate several programming strategies. Two programming strategies for a fairly common problem - a periodic tridiagonal solver - are developed and evaluated. Simple model calculations as well as timing results are presented to evaluate the various strategies. The particular tridiagonal solver evaluated is used in many computational fluid dynamic simulation codes. The feature that makes this algorithm unique is that these simulation codes usually require simultaneous solutions for multiple right-hand-sides (RHS) of the system of equations. Each RHS solutions is independent and thus can be computed in parallel. Thus a Gaussian elimination type algorithm can be used in a parallel computation and the more complicated approaches such as cyclic reduction are not required. The two strategies are a transpose strategy and a distributed solver strategy. For the transpose strategy, the data is moved so that a subset of all the RHS problems is solved on each of the several processors. This usually requires significant data movement between processor memories across a network. The second strategy attempts to have the algorithm allow the data across processor boundaries in a chained manner. This usually requires significantly less data movement. An approach to accomplish this second strategy in a near-perfect load-balanced manner is developed. In addition, an algorithm will be shown to directly transform a sequential Gaussian elimination type algorithm into the parallel chained, load-balanced algorithm.

Eidson, T. M.↗

NASA NTRS2005

PlanWorks: A Debugging Environment for Constraint Based Planning Systems

Numerous planning and scheduling systems employ underlying constraint reasoning systems. Debugging such systems involves the search for errors in model rules, constraint reasoning algorithms, search heuristics, and the problem instance (initial state and goals). In order to effectively find such problems, users must see why each state or action is in a plan by tracking causal chains back to part of the initial problem instance. They must be able to visualize complex relationships among many different entities and distinguish between those entities easily. For example, a variable can be in the scope of several constraints, as well as part of a state or activity in a plan; the activity can arise as a consequence of another activity and a model rule. Finally, they must be able to track each logical inference made during planning. We have developed PlanWorks, a comprehensive system for debugging constraint-based planning and scheduling systems. PlanWorks assumes a strong transaction model of the entire planning process, including adding and removing parts of the constraint network, variable assignment, and constraint propagation. A planner logs all transactions to a relational database that is tailored to support queries for of specialized views to display different forms of data (e.g. constraints, activities, resources, and causal links). PlanWorks was specifically developed for the Extensible Universal Remote Operations Planning Architecture (EUROPA(sub 2)) developed at NASA, but the underlying principles behind PlanWorks make it useful for many constraint-based planning systems. The paper is organized as follows. We first describe some fundamentals of EUROPA(sub 2). We then describe PlanWorks' principal components. We then discuss each component in detail, and then describe inter-component navigation features. We close with a discussion of how PlanWorks is used to find model flaws.

Daley, Patrick↗

NASA NTRS2015

Spinoff 2015

Topics covered include: 3D Endoscope to Boost Safety, Cut Cost of Surgery; Audio App Brings a Better Night's Sleep Liquid Cooling Technology Increases Exercise Efficiency; Algae-Derived Dietary Ingredients Nourish Animals; Space Grant Research Launches Rehabilitation Chair; Vision Trainer Teaches Focusing Techniques at Home; Aircraft Geared Architecture Reduces Fuel Cost and Noise; Ubiquitous Supercritical Wing Design Cuts Billions in Fuel Costs; Flight Controller Software Protects Lightweight Flexible Aircraft; Cabin Pressure Monitors Notify Pilots to Save Lives; Ionospheric Mapping Software Ensures Accuracy of Pilots' GPS; Water Mapping Technology Rebuilds Lives in Arid Regions; Shock Absorbers Save Structures and Lives during Earthquakes; Software Facilitates Sharing of Water Quality Data Worldwide; Underwater Adhesives Retrofit Pipelines with Advanced Sensors; Laser Imaging Video Camera Sees through Fire, Fog, Smoke; 3D Lasers Increase Efficiency, Safety of Moving Machines; Air Revitalization System Enables Excursions to the Stratosphere; Magnetic Fluids Deliver Better Speaker Sound Quality; Bioreactor Yields Extracts for Skin Cream; Private Astronaut Training Prepares Commercial Crews of Tomorrow; Activity Monitors Help Users Get Optimum Sun Exposure; LEDs Illuminate Bulbs for Better Sleep, Wake Cycles; Charged Particles Kill Pathogens and Round Up Dust; Balance Devices Train Golfers for a Consistent Swing; Landsat Imagery Enables Global Studies of Surface Trends; Ruggedized Spectrometers Are Built for Tough Jobs; Gas Conversion Systems Reclaim Fuel for Industry; Remote Sensing Technologies Mitigate Drought; Satellite Data Inform Forecasts of Crop Growth; Probes Measure Gases for Environmental Research; Cloud Computing Technologies Facilitate Earth Research; Software Cuts Homebuilding Costs, Increases Energy Efficiency; Portable Planetariums Teach Science; Schedule Analysis Software Saves Time for Project Planners; Sound Modeling Simplifies Vehicle Noise Management; Custom 3D Printers Revolutionize Space Supply Chain; Improved Calibration Shows Images' True Colors; Micromachined Parts Advance Medicine, Astrophysics, and More; Metalworking Techniques Unlock a Unique Alloy; Low-Cost Sensors Deliver Nanometer-Accurate Measurements; Electrical Monitoring Devices Save on Time and Cost; Dry Lubricant Smooths the Way for Space Travel, Industry; and Compact Vapor Chamber Cools Critical Components.

Source record↗

NASA NTRSDate not supplied

Adding GPU Support to the Markov Chain Monte Carlo Code Catmip

In geophysics, we are confronted with many under-determined inverse problems. For example, all of our observations of earthquakes are made at the Earth’s surface. So, when we try to infer how slip during an earthquake evolves in space and time, we find that there are many potential slip histories that are consistent with our limited observations and our understanding of earthquake physics. One way to approach these problems is with Bayesian analysis which allows us to infer the ensemble of all potential slip models that satisfy the observations and our prior knowledge of earthquake physics. In Bayesian analysis, our prior knowledge is known as the prior probability density function or prior PDF, the fit to the data is known as the data likelihood, and the target PDF that satisfies both the prior PDF and data likelihood is known as the posterior PDF. However, simulating the posterior PDF typically requires using Markov Chain Monte Carlo (MCMC) to draw tens of billions of random realizations of earthquake slip models, which may not be computationally feasible. To make this and similar geophysical inversions computationally tractable, we developed the Cascading Adaptive Transitional Metropolis In Parallel (CATMIP) algorithm. CATMIP is an efficient parallel Markov Chain Monte Carlo (MCMC) sampler that is used for model fitting and uncertainty quantification in geophysics. Example use cases are earthquake rupture modeling, determining mineral composition on Mars, reconstructing the history of ocean salinity, and historical earthquake relocation. CATMIP employs many parallel instances of the Metropolis algorithm for sampling in a transitioning framework. Transitioning is a process in which a set of random samples at equilibrium with a known probability density function (PDF) are used as seeds for the Markov chains to sample successive target PDFs that incrementally move the distribution from the starting seeds to the final desired PDF that describes the relative plausibility of potential values for the model parameters. The algorithm is implemented as a Master-Worker model employing MPI for communication. The worker processes are loosely coupled with global parameters periodically optimized by the master process. This provides a very high amount of parallelism with little communication between updates. During the presentation we will discuss the history of the algorithm and elaborate the earthquake rupture modeling use case for the CATMIP package. Our first step toward GPU optimization was to optimize the code for the CPU. CPU profiling revealed that most of the compute time is spent in calls to level 2 BLAS routines and calls to GSL random number generators. We revised the algorithm to employ level 3 BLAS routines instead. In our presentation we will describe how this was accomplished. Adding GPU support to CATMIP consisted mostly of replacing the calls to GSL with calls to GPU vendor-provided library routines. A small number of loops were directly implemented in CUDA. In the presentation will provide implementation details. Finally, we will discuss methods for profiling and opportunities for further optimizing GPU execution. By creating a code with the flexibility to run on either a CPU or GPU architecture, CATMIP can be used on systems ranging from large CPU-based HPC environments to single servers with GPU acceleration and everything in between.

HECC↗