考虑换乘点修建的多模式离散网络设计研究

出行者进行一次完整的出行往往会通过多种交通模式来实现, 其中停车换乘有效衔接了小汽车出行模式向公共交通模式的转换, 在此过程中需要在城市中设置特定区域或设施以完成停车换乘。文章提出了一个考虑换乘点修建的多模式网络设计问题(network design problem, NDP), 构建了一种双层规划模型, 上层模型从决策者角度考虑换乘停车点修建规划, 下层模型为多模式网络下的模式划分和交通分配问题。通过提出基于代理模型的算法来求解上述双层规划模型, 结合自适应平均法(self-regulated averaging method, SRAM) 和梯度投影(gradient projection, GP) 算法解决下层模式划分与交通分配问题。最后采用 Sioux Falls 网络进行数值算例分析, 验证了算法和模型的有效性。结果表明, 停车换乘点的合理修建能够有效降低城市交通网络的系统总费用, 提出的基于代理模型算法相较遗传算法在效率和精度上更高。

Travelers often employ various modes of transportation to accomplish a complete trip. Among them, park-and-ride facilitates the transition from car travel to public transportation systems. This process necessitates the establishment of designated areas or facilities within the city to enable park-and-ride transfers. Consequently, this paper proposes a multimodal network design problem (NDP) that considers the construction of transfer nodes. A bilevel programming model is constructed, where the upper-level model considers the decision-makers' perspective in planning the placement of park-and-ride facilities, while the lower-level model considers the mode division and traffic allocation problems under the multimodal network. A surrogate-model-based algorithm is presented to solve the aforementioned bilevel programming model, incorporating the self-regulated averaging method (SRAM) and gradient projection (GP) algorithm to address mode splitting and traffic assignment in the lower level. Finally, numerical experiments are conducted on the Sioux Falls network to validate the effectiveness of the algorithm and model. The experimental results demonstrate that the rational placement of park-and-ride facilities can effectively reduce the overall system cost of the urban transportation network. The proposed surrogate-model-based algorithm outperforms genetic algorithms in terms of both efficiency and accuracy.