A change severity degree-based dynamic multi-objective optimization algorithm with adaptive response strategy

Abstract

Many real-world optimization problems are dynamic by nature, exhibiting temporal variations in objective functions, constraints, and parameters. These problems present significant challenges for algorithm convergence and diversity, as they require the ability to adapt appropriately to new environments. To address these challenges, a Dynamic Multi-objective Particle Swarm Optimization algorithm with an Adaptive Response Strategy (DMOPSO-ARS) is proposed in this paper. The DMOPSO-ARS possesses the capability to detect the degree of change and apply a suitable response strategy accordingly. Specifically, the change response strategy encompasses initialization-based prediction and elite-based learning methods, designed to speed up convergence speed and enhance algorithm diversity in the face of high/low severity environmental changes. To assess the effectiveness of the DMOPSO-ARS, we select the standard CEC 2018 benchmark, noisy test functions, and the recently released OL-DOP 2022 benchmark. Empirical studies indicate the robustness of the DMOPSO-ARS in tracking evolving solutions over time, showcasing its significant superiority compared to state-of-the-art methods.