A Particle Swarm Optimization with Dynamic Strategy for Multi-Modal Multi-Objective Location Optimization Problem

Abstract

Multi-modal multi-objective optimization problems (MMOPs) are increasing popularity recently. They show a many-to-one mapping throughout the spaces and are made up of several conflicting objective functions that must be optimized simultaneously. Thus, We propose a particle swarm optimization with a dynamic strategy to improve search efficiency for solving MMOPs. Sub-populations are formed based on the dynamic radius. Next, each individual will update its position based on both the center solution of its sub-population and one of its own personal best positions. The effectiveness of PSO-DN is demonstrated on the location optimization problem generated from the real-world map. Compared to four state-of-the-art algorithms, PSO-DN achieves superior results for MMOPs. Both the number of Pareto-optimal sets and the Hv in the objective space demonstrate this superiority.