Abstract: Transshipment hub ports typically handle high-density water-to-water transshipment,while limited port resources pose significant challenges for efficient berth scheduling. To address this issue,a berth scheduling approach integrating multiple transshipment modes is proposed by incorporating a double-berthing operation,which allows vessels to berth side-by-side at the same berth and conduct direct transshipment via quay cranes. The double-berthing operation is further integrated with indirect transshipment and improved direct transshipment to form a coordinated transshipment framework. A mixed-integer programming model is formulated to jointly determine berth allocation and transshipment mode selection,with the objective of minimizing transshipment costs and vessel-related operational costs. The model incorporates practical constraints including transshipment time windows,vessel matching relationships,and quay crane width limitations. To solve the resulting complex optimization problem,an Adaptive Large Neighborhood Search(ALNS) algorithm is developed.Multiple destroy and repair operators together with an adaptive weight updating mechanism are designed to enhance the exploration capability and computational efficiency of the algorithm. Computational experiments of different scales are conducted,and the proposed algorithm is compared with the Gurobi solver,Genetic Algorithm(GA),and Simulated Annealing(SA). In addition,weight sensitivity analysis is performed to evaluate the robustness of the model under different cost weight settings. The applicability of the proposed approach is further verified using real arrival data from feeder berths at Shanghai Yangshan Port. The results show that,compared with traditional scheduling schemes,the proposed approach significantly reduces vessel waiting time and overall completion time while substantially decreasing truck transportation and yard storage costs. Specifically,the total vessel waiting time is reduced by 63.48%,the total operational time by 5.73%,and the overall operational cost by 65.09%. Furthermore,the improved ALNS achieves solutions within 3% of those obtained by Gurobi for small-and medium-sized instances,and obtains high-quality solutions within shorter computation times than GA and SA for large-scale instances,demonstrating strong robustness and computational efficiency.