Multi-Objective Artificial Physics Optimization Algorithm Based on Population Dynamic Clustering Strategy

Aiming at the problems that the multi-objective artificial physics optimization algorithm (MOAPO) is prone to fall into local optima and has poor distribution, a multi-objective artificial physics optimization algorithm based on the population dynamic clustering strategy (DCMOAPO) is proposed. Firstly, by introducing the population dynamic clustering mechanism, the population is divided into multiple subpopulations, and parallel exploration is carried out within different search regions to enhance the local search ability while maintaining the global diversity of the swarm. In addition, an individual attraction mechanism for the archive set is designed. Potential individuals are selected as guiding points to prompt the population to converge towards the region of better solutions. Finally, in terms of the archive set update, combining the non-dominated sorting and the KNN density estimation method, the archive set is updated dynamically to further improve the diversity and the uniform distribution of the solution set. Comparative experiments with a number of typical multi-objective optimization algorithms across 12 benchmark functions. Results show this algorithm has notable competitive edges and effectiveness in convergence and solution diversity.