A meta‐level evolutionary strategy for many‐criteria design: Application to improving tracking filters

We present a novel meta‐level heuristic algorithm for multi‐criteria search. It focuses on dynamically adapting the optimization criteria through the set of active objectives instead of using the evolutionary strategy (ES) parameters as other meta‐level approaches do. The meta‐level ES dynamically searches for the subset of objectives that achieves the best global performance. It assumes that the active subset can represent the real structure of the trade‐off surface and consider all objectives at the same time as a pure multi‐objective evolutionary approach (MOEA) would do. We have successfully applied this heuristic to improve the efficiency of tracking filters design, a real‐world problem requiring effective and fast optimization techniques. Our approach yields competitive results and drastically reduces the computational cost. The results show an important advantage in efficiency with respect to previous conventional approaches for applying evolutionary algorithms (EA) to the same design problem. The proposed technique can be applied to real‐world problems with a high number of active dependent objectives, a frequent occurrence in engineering design.