Self-adaptive opposition-based differential evolution with subpopulation strategy for numerical and engineering optimization problems

Abstract

Opposition-based differential evolution (ODE) is a well-known DE variant that employs opposition-based learning (OBL) to accelerate the convergence speed. However, the existing OBL variants are population-based, which causes many shortcomings. The value of the jumping rate is not self-adaptively adjusted, so the algorithm easily traps into local optima. The population-based OBL wastes fitness evaluations when the algorithm converges to sub-optimal. In this paper, we proposed a novel OBL called subpopulation-based OBL (SPOBL) with a self-adaptive parameter control strategy. In SPOBL, the jumping rate acts on the individual, and the subpopulation is selected according to the individual’s jumping rate. In the self-adaptive parameter control strategy, the surviving individual’s jumping rate in each iteration will participate in the self-adaptive process. A generalized Lehmer mean is introduced to achieve an equilibrium between exploration and exploitation. We used DE and advanced DE variants combined with SPOBL to verify performance. The results of performance are evaluated on the CEC 2017 and CEC 2020 test suites. The SPOBL shows better performance compared to other OBL variants in terms of benchmark functions as well as real-world constrained optimization problems.