Ternary Historical Memory-Based Robust Clustered Particle Swarm Optimization for Dynamic Berth Allocation and Crane Assignment Problem

The berth allocation and crane assignment problem (BACAP) is a key challenge in port logistics, particularly under dynamic and uncertain vessel arrival conditions. To address the limitations of existing methods in handling large-scale and high-disturbance scenarios, this paper proposes a novel optimization framework: Ternary Historical Memory-based Robust Clustered Particle Swarm Optimization (THM-RCPSO). In this method, the initial particle swarm is divided into multiple clusters, each conducting local searches to identify regional optima. These clusters then exchange information to iteratively refine the global best solution. A ternary historical memory mechanism further enhances the optimization by recording and comparing the best solutions from three different strategies, ensuring guidance from historical performance during exploration. Experimental evaluations on 25 dynamic BACAP benchmark instances show that THM-RCPSO achieves the lowest average vessel dwell time in 22 out of 25 cases, with the lowest overall average rank among five tested algorithms. Specifically, it demonstrates significant advantages on large-scale instances with 150 vessels, where it consistently outperforms competing methods such as HRBA, ACO, and GAMCS in both solution quality and robustness. These results confirm THM-RCPSO’s strong capability in solving dynamic and large-scale DBACAP scenarios with high disturbance levels.