Products

• 

  Autonomous Exploration for Gathering Increased Science (AEGIS)

  The AEGIS intelligent targeting system allows a Mars rover to detect, prioritize, select, and observe scientific targets on its own, without waiting for input from humans on Earth. AEGIS collects an image with a camera aboard the rover, analyzes the image to identify geological targets, chooses the target that best matches the parameters specified by operators, and then points a remote-sensing instrument at the selected target(s), all autonomously. First deployed on the MER rover Opportunity, AEGIS is currently used on the MSL mission to select targets for the Curiosity rover’s ChemCam instrument, a remote laser-induced breakdown spectrometer. While the majority of targets are still selected by humans, AEGIS allows the rover to continue its geochemical survey during periods when Earth cannot be in the loop, and has contributed measurements on hundreds of targets, and become a routine part of how we explore Mars.
• 

  Basic Instrumentation Tool for Fault Localized Injection of Probabilistic SEUs (BITFLIPS)

  BITFLIPS is a software simulator for injecting single event upsets (SEUs) into a running computer program. The software is written as a plugin extension module of the open source Valgrind debugging and profiling tool. BITFLIPS can inject SEUs into any program which can be run on the Linux operating system, without modifying the original program or requiring access to the program source code. If access to the original program source code is available, BITFLIPS can offer fine-grained control over exactly when and which areas of memory (program variables) may be subjected to SEUs. SEU injection rate is controlled by specifying either a fault rate based on memory size and radiation exposure time in terms of SEUs per kilobyte per instruction. BITFLIPS also has the capability to log each SEU it injects and, if program source code is available, report the magnitude of the SEU on the program variable.
• 

  COmplex Data EXplorer (CODEX)

  Modern science datasets may have 500+ simultaneous measurements at each of millions of time samples. Scientists would often like to look through the record and discover not only expected trends but ones they did not initially guess. The optimal environment for this rapid exploration large data would be one where visualizations were clear, interactive, and responsive, permitting the investigator to “play” with the data and gain rapid insight, falsify hypothesis, and make discoveries. Machine Learning has proven invaluable in providing some of these key data insights, but to do so in a statistically robust and reliable manner requires a data science professional and a lot of custom Python code, losing any sense of interaction and play. CODEX addresses these concerns by providing a desktop-like environment with standard scientific graph types that are robust to rapid, powerful exploration.
• 

  Data Ordering Genetic Optimization (DOGO)

  DOGO was developed on the Orbiting Carbon Observatory 2 (OCO-2) mission to identify features that predict when the received CO2 goes awry. It was later adopted to advise science users on the trustability of individual observations in the mission record via the Warn Level formula: a simple prescription that assigns each observation an integer from 0 to 9 estimating how much confounding forces are present. DOGO's greatest significance is its ability to predict data trust without the need for ground truth data, instead using a series of expert-derived sanity checks on the data to narrow down on the best and worse behaved observations. DOGO is currently applying to AMMOS for implementation funds to spread its technology to other future missions.
• 

  Distributed Optimization of Machine Learning Incorporating Nested Evaluation (DOMINE)

  DOMINE is a software tool that enables large-scale exploration and tuning of competing machine learning technologies in a fully robust, cross-validated, and distributed sense including feature selection. DOMINE asks users to specific their desired datasets, labels, and classifiers to enable a full and efficient diagnosis which technologies, hyperparameter settings, and feature settings perform best. Our system, by design, incorporates nested cross validation to ensure publication-quality results with statistically robust, defensible decisions for model optimality. All of this is accomplished on top of a client-server interface that allows for large-scale, distributed computation for quicker results.
• 

  Tuning Optimizing Genetic Algorithm (TOGA)

  The Tuning Optimizing Genetic Algorithm (TOGA) was developed on the Ocean Worlds Life Surveyor project to perform black-box optimization on software with large numbers of tunable parameters. In addition to featuring highly configurable and transparent mutators, crossover functions, population controls, and performance metrics, TOGA is able to take full advantage of high performance distributed computing with its client-server architecture for fast convergence. TOGA provides optimization for both classifier selection and hyperparameter tuning, in addition to supporting sensitivity studies and tradeoff analysis between configuration options.