A co-evolutionary meta-heuristic framework for dynamic constrained optimization problems

Abstract Dynamic constrained optimization problems (DCOPs) are optimization problems where both the problem landscape and the problem constraints change over time, or either the problem landscape or the constraints change. Although DCOPs represent the super-set of optimization problems, relatively little is understood about these problems due to the complexity added to the optimization process and few meta-heuristics exist for DCOPs. This paper proposes a co-evolutionary meta-heuristic framework to allow for easy integration of existing dynamic meta-heuristics developed to solve box-constrained dynamic optimization problems only, into the framework to produce new co-evolutionary versions of these dynamic meta-heuristics to solve various classes of DCOPs. The paper analyzes the performance of the resulting co-evolutionary versions of these existing dynamic meta-heuristics on a comprehensive set of DCOP benchmark problems, and shows that the performance of these dynamic co-evolutionary algorithms are the best performing among a number of evaluated meta-heuristics.