摘要：

随着我国城镇化进程的不断推进，城市人口和车辆数量持续增加，大宗货物运输导致的空气污染与交通拥堵问题已成为影响居民生活和制约城市发展的重要障碍。为缓解特大型城市地面道路资源紧张的现状，提出建立深层隧道运输与浅层管廊运输相结合的多层级地下物流网络，将部分地面货流转移至地下从而释放地面运输能力。从投入成本、资源利用效率等多个角度出发，考虑地下运输的多级结构特征，构建多层级地下物流网络优化设计的整数规划模型，得出合理的布局方案。根据问题特征，提出基于人工免疫与模拟退火的双层算法，通过均值偏移聚类算法对解空间进行分解，预先筛除部分明显不合理的决策方案, 在此约束基础上通过双层启发式算法进行节点选址和流量分配优化决策，以建设与运营成本最小为目标，通过多次迭代得出地下物流多层网络规划方案。数值实验和案例分析结果表明，与传统的遗传算法相比，该算法解决节点布局与网络规划问题的寻有能力提升了2%~7%，平均计算时长降低约50%，验证了所建模型的合理性。

关键词: 城市地下物流, 网络优化, 整数规划, 均值偏移聚类算法, 启发式算法

Abstract:

With the continuous advancement of urbanization in China, the urban population and the number of vehicles continue to increase. Air pollution and traffic congestion resulting from bulk cargo transportation have become significant obstacles to improving the living conditions of residents and urban development.To relieve the current shortage of road transportation resources in megalopolises, a multilevel underground logistics network combining deep tunnel transportation and shallow pipebag channel transportation is proposed. Part of the cargo fluence to was transferred underground to release the transport capacity. An integer programming model for the optimal design of the multilevel underground logistics network is formulated from the perspectives of cost and resource utilization efficiency.Based on the characteristics of the problem and the multilevel structural features of underground transportation, a two-layer heuristic algorithm based on simulated annealing and an immune algorithm is designed as follows.First, the solution space is decomposed using mean-shift clustering to remove unreasonable decisions.Subsequently, the two-layer heuristic algorithm is used to optimize the decision-making of node locations and traffic allocation.To minimize total construction and operating costs, a satisfactory underground logistics multilayer network planning scheme can be obtained through multiple iterations.The numerical experiment results and a case study demonstrate that the proposed algorithm outperforms the traditional genetic algorithm. It improves solution optimization by 2% to 7% for node layout and network planning problems while reducing computation time by approximately 50%, further validating the proposed model.

Key words: urban underground logistics, network optimization, integer programming, mean-shift clustering algorithm, heuristic algorithm