Genetic algorithms-based techniques for solving dynamic optimization problems with unknown active variables and boundaries

Abstract

In this paper, we consider a class of dynamic optimization problems in which the number of active variables and their boundaries vary as time passes (DOPUAVBs). We assume that such changes in different time periods are not known to decision makers due to certain internal and external factors. Here, we propose three variants of genetic algorithm to deal with a dynamic problem class. These proposed algorithms are compared with one another, as well as with a standard genetic algorithm based on the best of feasible generations and feasibility percentage. Experimental results and statistical tests clearly show the superiority of our proposed algorithms. Moreover, the proposed algorithm, which simultaneous addresses two sub-problems of such dynamic problems, shows superiority to other algorithms in most cases.