Dynamic ε-multilevel hierarchy constraint optimization with adaptive boundary constraint handling technology

Abstract

Real-world optimization problems are often difficult to solve because of the complexity of the objective function and the large number of constraints that accompany it. To solve such problems, we propose Adaptive Dynamic ε-Multilevel Hierarchy Constraint Optimization (εMHCO). Firstly, we propose the dynamic constraint tolerance factor ε which can change dynamically with the feasible ratio and the number of iterations in the current population. This ensures a reasonable proportion of virtual feasible solutions in the population. Secondly, we propose adaptive boundary constraint handling technology (ABCHT). It can reshape the current individual position adaptively according to the size of constraint violation and increase the diversity of the population. Finally, we propose multi-level hierarchy optimization, whose multiple population structure is beneficial to solve real-world constraint optimization problems (COPs). To validate and analyze the performance of εMHCO, numerical experiments are conducted on the latest real-world test suite CEC’2020, which contains a set of 57 real-world COPs, and compared with four state-of-the-art algorithms. The results show that εMHCO is significantly superior to, or at least comparable to the state-of-the-art algorithms in solving real-world COPs. Meanwhile, the effectiveness and feasibility of εMHCO are verified on the real-world problem of the pipeline inner detector speed control.