A Consensus-Based Particle Swarm Optimization Algorithm for Distributed Multi-Objective Optimization

Distributed optimization has been extensively studied in the past decades due to its wide applications. Most of them aim to optimizing single-objective function. However, in real-world applications and industrial production, it is often necessary to trade off between two or more objective functions, requiring algorithms capable of optimizing multiple objectives simultaneously. In view of this, a consensus theory and particle swarm-based distributed multi-objective optimization algorithm (CTPS-DMOA) is proposed, which can achieve a consensus Pareto optimal solution in a distributed manner. In CTPS-DMOA, a population is assigned to each local multi-objective function. Then, an average consensus-based particle communicating operator is introduced, allowing neighboring particles at the same layer to exchange useful information over a connected communication graph, ensuring that particles at the same layer can reach consensus on their positions. To address the evaluation issue of particles, a average consensus-based global evaluation operator is proposed to ensure a global evaluation of particles only with local information. Additionally, to enhance the diversity of solutions and output a more complete Pareto optimal solution set, an extreme solutions guided Pareto front stretching operator is introduced. Finally, the overall framework of CTPS-DMOA is presented, and a series of experiments validate the superiority of CTPS-DMOA in solving distributed multi-objective optimization problems.