Projects

Optimization R&D encompasses many areas of research in optimization including the development of algorithms for very expensive functions and large-scale problems. Areas included under this project include parallel optimization methods, asynchronous methods, model management methods, and interior point methods. In addition, we are working on object-oriented optimization toolkits including a C++ library, OPT++ and a Java Optimization Toolkit, JOT.

MAUI is a tool for rapidly developing a graphical user interface (GUI) for an application based on a high-level description of the structure and parameters of the application. MAUI is being used in several Sandia applications.

DDACE - A Distributed Design and Analysis of Computer Experiments. This project aims to develop algorithms for the design and analysis of computer experiments. Its current capabilities include the ability to generate points from various sampling techniques, to build response surfaces, and to do statistical analysis based on the computed function values.

Multilevel Methods for the solution of large-scale linear systems of equations is an active area of research within the department. We are developing new parallel multigrid methods for solving large-scale systems of linear equations arising from both time dependent and steady-state PDE's.

Fault Tolerant Algorithms are an essential ingredient for large-scale simulations that can run for days on parallel processors. These simulations frequently crash in the event of a node failure. The current research involves the development of fault tolerant algorithms for large heterogeneous cluster systems assembled from commodity parts.

DOE2000: ACTS, the Advanced Computational Testing and Simulation project, is tasked to build a set of software tools, known as the ACTS Toolkit, for high performance computing. The purpose of ACTS is to accelerate the adoption and use of advanced computational techniques and methods in mission-critical problems of the Department of Energy.

Stochastic Programming R&D includes a wide range of efforts inclusive of stochastic optimal control, chance-constrained programming, multi-stage programs and recourse problems. Our goal is not only to formulate stochastic programming problems that intelligently "incorporate" uncertain information, but also to develop robust and efficient stochastic programming methodologies.

Uncertainty Quantification involves characterization and quantification of uncertainty in a simulation. Current efforts involve the development of analyis tools that can be used in optimal control and design problems.