Guideline Identification for Optimization Under Uncertainty Through the Optimization of a Boomerang Trajectory

Abstract

Optimization under uncertainty (OUU) is a very important task for practitioners of engineering design optimization. In fact real–world problems are often affected by uncertainties of different kind. The search for robust optimal solutions is intrinsically multiobjective, being formulated as the search for the optimal performance while minimizing its variance. Thus, OUU should garner interest in the evolutionary multiobjective optimization community. It is a challenging topic, because, for instance, engineers have to deal with large scale or highly–constrained problems. The first issue affects the feasibility of the optimization itself, whereas the second affects the reliability of an optimal solution. In this paper, we address the OUU problem to validate a number of best practices through the application to a benchmark problem: the optimization of a boomerang launch parameters. To reduce the computational cost, we consider variable screening as a preliminary step before performing a stochastic optimization. For the latter we use a method recently proposed by the authors, which combines robustness and reliability assessments within a single optimization run.