Abstract
Gradient-based numerical optimization of complex engineering designs promises to produce better designs rapidly. However, such methods generally assume that the objective function and constraint functions are continuous, smooth, and defined everywhere. Unfortunately, realistic simulators tend to violate these assumptions. We present several artificial intelligence-based techniques for improving the numerical optimization of complex engineering designs in the presence of such pathologies in the simulators. We have tested the resulting system in several realistic engineering domains, and have found that using our techniques can greatly decrease the cost of design space search, and can also increase the quality of the resulting designs.