Evolutionary algorithms based multiobjective optimization techniques for intelligent systems design

We present evolutionary algorithm based multiobjective optimization techniques for intelligent systems design. Multiobjective optimization techniques are necessary in situations where the performance of a system is based on multiple, possibly conflicting objectives whose aggregation cannot be easily articulated. The evolutionary algorithms approach presented employs a search mechanism that treats each of the objectives independently, avoiding the objective aggregation step. A key feature of our techniques is that they output a set of solutions rather than a single solution. To demonstrate how our techniques can be used to support system design, we apply them to the task of designing a fuzzy control system. In the final part of the paper, we propose metrics for multiobjective optimization algorithm performance and techniques for employing them in the design an adaptation of evolutionary algorithm based multiobjective optimization.