A SEARCH SPACE TOOLKIT - SST

The Search Space Toolkit (SST) is a suite of tools for investigating t he properties of the continuous search spaces which arise in designing complex engineering artifacts whose evaluation requires significant c omputation by a numerical simulator. SST has been developed as part of NDA, a computational environment for (semi-)automated design of jet e ngine exhaust nozzles for supersonic aircraft which resulted from a co llaboration between computer scientists at Rutgers University and desi gn engineers at General Electric and Lockheed. Though the design space s for this sort of engineering artifact are mainly continuous, they ty pically include features such as unevaluable points, multiple local op tima, and large derivatives which cause difficulties for standard nume rical optimization methods. The search spaces which SST explores also differ significantly from the discrete search spaces that typically ar ise in artificial intelligence research, and properly searching such s paces requires a synergistic combination of numerical methods and AI t echniques and is a fundamental AI research area. By promoting the desi gn space to be a first class entity, rather than a ''black box'' burie d in the interface between an (unconstrained) optimizer and a simulato r, SST allows a more principled approach to automated design.