A new subpopulation model for evolutionary multimodal optimization

A new evolutionary technique for multimodal optimization called roaming optimization (RO) is presented. Multiple optima are detected using subpopulations evolving in isolation. A new search operator called migration is introduced. A stability measure is defined for subpopulations by which they are characterized as stable or unstable. Stable subpopulations are considered to contain local optima. An external population called the archive is used to store the optima detected. After a number of generations the archive contains all local optima. Experimental results prove the efficiency of the algorithm.