Project 4-4: High Performance Optimization Library
Principal Investigators: Walter Murray, Michael Saunders (Stanford)
(This project ran from 2007 to 2009. Project 1-7 is an outgrowth of this work.)

Optimization is a vital part of most branches of computational science and engineering. General-purpose optimization packages are constantly finding new applications in aerospace, economics, geophysics, manufacturing, and many more areas. In aerospace, shape optimization and trajectory optimization for aircraft and spacecraft are important to all branches of the Department of Defense.

AHPCRC researchers are in the forefront of designing and implementing effective optimization software. The goal of this project is to enhance existing packages by incorporating high performance computing techniques such as parallel linear-system solvers, and to assist other AHPCRC projects in need of optimization techniques.

An example is project 1–3, where micro aerial vehicles (MAVs) are designed to maximize lift and minimize drag. This requires interaction between fluid-flow partial differential equation solvers and optimization solvers requesting a sequence of solutions to partial differential equations. The model formulation should take advantage of the fact that the optimal solution values for neighboring variables tend to be similar (an assumption not made in general-purpose optimization solvers).

Another aim is to apply efficient continuous optimization methods to large problems involving integer or discrete-valued variables.