Structured Controller Design using Evolution Strategies: A Computational Analysis

A set of evolution strategies is implemented in order to find the optimal tuning of a structured controller. The structured controller design problem is a non-convex optimization problem. The usual solution technique involves the use of the expansive branch and bound global search algorithm. Results based on the Kruskal-Wallis statistic and Dunn-Sidak post test showed that the evolution strategies were able to find the global optimum with comparable solution quality, less overhead and considerable gain in computation time. Surprisingly, statistical tests revealed that the simple two member evolution strategy´s performance is not significantly different from other evolution strategies