A Tool for Computationally Steering Optimization Software

New methods for improving optimization software using random techniques and visualization were proposed and shown feasible in Phase I of this effort using a prototype GUI implemented with OpenGL. The Phase II effort presents five technical tasks necessary to put these ideas into practice. The first implements new algorithms for random methods using number theory, finite field theory, and advanced algebra. The second implements new algorithms for approximating expensive functions with ones less expensive to evaluate (surrogates). The third implements new ideas for the interactive visualization of mathematical data. The fourth is a spin off of the prototype mentioned above producing an interactive environment for experimenting with optimization software design. The fifth, also a descendent of the prototype tool mentioned above, implements a GUI for ``computationally steering'' optimization software. The products of this effort will aid the Air Force in solving realistic aerospace systems optimization problems.