Multi-guide Particle Swarm Optimisation for Dynamic Multi-objective Optimisation Problems

Abstract

This study investigates the suitability of, and adapts, the multi-guide particle swarm optimisation (MGPSO) algorithm for dynamic multi-objective optimisation problems (DMOPs). The MGPSO is a multi-swarm approach, originally developed for static multi-objective optimisation problems (SMOPs), where each subswarm optimises one of the objectives. It uses a bounded archive that is based on a crowding distance archive implementation. Compared to static optimization problems, DMOPs pose a challenge for meta-heuristics because there is more than one objective to optimise, and the location of the Pareto-optimal set (POS) and the Pareto-optimal front (POF) can change over time. To efficiently track the changing POF in DMOPs using MGPSO, six archive management update approaches, eight archive balance coefficient initialization strategies, and six quantum particle swarm optimisation (QPSO) variants are proposed. To evaluate the adapted MGPSO for DMOPs, a total of twenty-nine well-known benchmark functions and six performance measures were implemented. Three experiments were run against five different environment types with varying temporal and spatial severities. The best strategies from each experiment were then compared with the other dynamic multi-objective optimisation algorithms (DMOAs). An extensive empirical analysis shows that the adapted MGPSO achieves very competitive, and often better, performance compared to existing DMOAs.