A Proposal of a Multi-Objective Compact Particle Swarm Optimizer

Abstract

Throughout the years, several bio-inspired meta-heuristics have been proposed to solve multi-objective problems. Nevertheless, most of the current metaheuristics are not suitable for applications having limited resources (e.g., limited available memory or computationally expensive objective function evaluations). In recent years, a wide variety of metaheuristics have been proposed that employ a statistical representation of the population through a probabilities vector. These are the so-called compact metaheuristics. Several metaheuristics of the state of the art have used a statistical representation to reduce the amount of memory required to be implemented in devices with limited computing resources. This paper presents a compact metaheuristic based on a particle swarm optimizer (PSO) for solving continuous and unconstrained multi-objective optimization problems. Our proposed approach is compared with respect to two multi-objective particle swarm optimizers (MOPSOs) and one compact multi-objective evolutionary algorithm (MOEA). The results indicate that our proposed approach is competitive with respect to the other MOPSOs and is able to outperform the compact MOEA used in our comparative study in most of the test problems adopted.