Next generation electromagnetic optimization with the covariance matrix adaptation evolutionary strategy

Abstract

Classical evolutionary strategies such as the genetic algorithm and particle swarm technique have long been the most called upon methods for optimization of electromagnetic design problems. Due to their capability for robust global search and their ease of implementation, they have been fruitfully applied to the design of antennas, arrays, frequency selective surfaces, metamaterials and other electromagnetic devices. Since then, many new optimization techniques have been developed that often allow more complex design problems to be tackled, or reduce the time needed to optimize the problems of the past. One algorithm found particularly effective is the covariance matrix adaptation evolutionary strategy (CMA-ES). The operation of CMA-ES will be covered in detail here. Additionally, the powerful performance of the technique when confronted with several different design problems and test functions will be demonstrated.