Radial Basis Function Neural Networks-Based Surrogate Model for Dynamic Multi-Objective Optimization

This paper introduces a novel surrogate modeldriven strategy to solve dynamic multi-objective optimization problems (DMOPs) with time-varying objective functions. This strategy holds promise in addressing high-dimensional nonlinear problems, particularly those with high-dimensional decision variables, while adapting to dynamic environmental changes. Our method employs an surrogate model, Radial Basis Function (RBF) neural network, to learn the nonlinear relationships between the decision space and the objective space. This model is then used to estimate population fitness, enabling the Dualoffspring evolution process. Furthermore, by leveraging information from historically evolved populations through Dual-offspring evolution, our approach enhances the prediction of future population distributions. Comparative evaluations on benchmark tests demonstrate that our method outperforms other algorithms in most test scenarios.