An archive maintenance scheme for finding robust solutions

This paper presents an archive maintenance scheme that can be used within an evaluation scheme for finding robust optima when dealing with expensive objective functions. This archive maintenance scheme aims to select the additional sampling points such that locally well-spread distributions of archive points will be generated. By doing so, the archive will contain better predictive information about the robustness of candidate solutions. Experiments on 10D test problems show that this scheme can be used for accurate local search for robust solutions.