An Improved Heterogeneous Comprehensive Learning Symbiotic Organism Search for Optimization Problems

Abstract

The Symbiotic Organism Search (SOS) algorithm, and most of its variants, exhibit the same search behaviour. That is, each organism or individual in the ecosystem applies the same update rules according to 3 phases: mutualism, commensalism, and parasitism. By applying this strategy, an organism is capable of crippling the search effort of another organism. To overcome this problem, this paper proposes a Heterogeneous Comprehensive Learning SOS (HCLSOS) which divides the population into two distinct subpopulations namely: the exploration and exploitation subpopulations. HCLSOS allows each organism to follow one of two search behaviour according to into its subpopulation: whether to explore or exploit. This addresses the problem of balancing exploration and exploitation in SOS. The proposed algorithm employs a comprehensive learning strategy for the exploration group to generate a new type of mutual vector called the multispecies mutual vector capable of preserving organisms' diversity and discouraging premature convergence. Information exchange between the two groups is unidirectional and managed through a random elite learning strategy. Through this collaboration, HCLSOS can effectively evolve organisms to explore the search space and properly exploit the discovered optimal regions. The study tested HCLSOS on 23 benchmark functions, CEC2014, CEC2017, and the recent CEC 2022 test suites. The outcome of HCLSOS is compared with 15 state-of-the-art algorithms, and the results obtained showed that HCLSOS can attain competitive or even better results. Furthermore, we applied HCLSOS to solve 3 constrained engineering problems and to design an efficient frequency control of a two-area islanded microgrid system.