Abstract
Dynamic optimization holds promise to solve real world problems that require adaptation to dynamic environments. The challenge is to track optima in an ever changing landscape. This paper describes a new computational intelligence approach to dynamic optimization termed as wind driven dynamic optimization (WD2O). Basically, it relies on an enhanced Multi-Region Modified Wind Driven Optimization (MR-MWDO) model and exhibits four main features. First, a multi-region approach is used to classify regions of the search space into promising and non-promising areas with accordance to low and high pressure regions in the natural model. Second, it uses an effective collision avoidance strategy to prevent collision between sub-populations. Third, it allows cost effective change detection. Fourth, it maintains two types of populations in order to achieve better balanced search. The proposed WD2O has been successfully applied to Moving Peaks Benchmark (MPB) problem. An extensive experimental study has shown that WD2O outperforms significantly the first prototype MR-MWDO. Furthermore, it has shown very competitive results compared to state of the art methods and has achieved the best performance for high dimensional problems while keeping an appreciable time complexity.