A Comparative Study of Multi-Guide Particle Swarm Optimization Topologies in Dynamic Multi-Objective Environments

Multi-objective optimization problems (MOPs) contain two or three objectives which need to be optimized simultaneously. Rather than having a single optimal solution, MOPs have a set of optimal trade-off solutions. Optimization is made more difficult in the case of dynamic MOPs (DMOPs). Multi-guide particle swarm optimization (MGPSO) has been introduced as a method to optimize static MOPs and DMOPs by optimizing each objective with its own sub-swarm. This paper looks at the impact of using different MGPSO swarm topologies to determine which is the best choice when using MGPSO to solve DMOPs. A benchmark set of 29 dynamic benchmark functions were used to evaluate the performance of each topology. Six performance measures were used for comparison. The results indicate that the ring topology is best suited to the dynamic environments tested here in general, especially Type I and Type II DMOPs. The wheel topology was found to perform best in Type III DMOPs.