Abstract
This paper proposes a multi-population-based multi-objective evolutionary algorithm (MP-MOEA) for solving complex maritime inventory routing problems, aiming to simultaneously minimize transportation costs and greenhouse gas emissions. To maintain population diversity, this paper employs a multi-population-based initialization operator to generate multiple populations containing solutions with varying numbers of vessels. Additionally, the proposed initialization operator utilizes a dynamic three-level encoding strategy, which significantly reduces the dimensionality of decision variables and lowers the complexity of encoding and decoding compared to traditional fixed-length encoding. To address the complex constraints of the studied maritime inventory routing problem, an individual modification operator is designed to improve solution feasibility. Furthermore, to accelerate population convergence and expand the search range, a hybrid crossover operator and a contribution-based mutation operator are proposed to balance the convergence and diversity in MP-MOEA. In this paper, the proposed MP-MOEA is compared with five state-of-the-art multi-objective evolutionary algorithms, including NSGAII, BiCo, MSCEA, TSTI, and AGEMOEAII, on maritime inventory routing problems of three different scales. The experimental results indicate that the solutions provided by the MP-MOEA outperform those of the other compared algorithms in addressing different problem instances.