Robust Optimal Design Under Standard Crowding Differential Evolution Framework

In practical optimization problems, it is often desired that a solution is not only of high performance, but also of high robustness, for example, robust to fabrication tolerances. The robust optimal design problem is studied under a standard crowding differential evolution framework in this paper, where the robustness of a solution is indicated by the objective in the worst case of this solution. To evaluate the robust objective function, a simple evolutionary strategy is used, other than the traditional Monte-Carlo simulation. Numerical results indicate that the proposed method can provide faster convergence rate and more accurate solution than Monte-Carlo simulation. This method is then applied to the practical holographic grating design, and achieves satisfactory optimization results