A general formulation for multi-modal dynamic traffic assignment considering multi-class vehicles, public transit and parking
A general formulation for multi-modal dynamic traffic assignment considering multi-class vehicles, public transit and parking
复制标题
DOI：
10.1016/j.trc.2019.05.011
复制
发表时间：
2019-07-01
影响因子：
8.3
通讯作者：
Qian, Zhen (Sean)
中科院分区：
文献类型：
--
作者：
Pi, Xidong;Ma, Wei;Qian, Zhen (Sean)
现代城市交通系统中的出行行为和出行成本受到许多方面的影响，包括道路上的异构交通（私家车、货运卡车、公共汽车等）、目的地附近的停车位以及网络中可用的出行方式，如单人驾驶、拼车、网约车、公共交通和停车换乘。管理这样一个复杂的多式联运系统需要一个整体的交通网络流建模框架，包括客流和车流。本文在明确考虑多类别车辆、停车设施和多种出行方式的情况下，建立并求解了一般多模式网络的时空客流和车辆流。车辆流，即汽车、卡车和公共汽车，被整合到一个整体动态网络加载（DNL）模型中。利用多层嵌套logit模型对乘客出行方式和路线选择的需求行为进行了封装。我们将多模态动态用户平衡（MMDUE）转化为变分不等式（VI）问题。本文从KKT条件出发，推导了一种基于梯度投影的起点-终点水平的简单流解，并证明了它可以有效地解决大规模网络上的VI问题。在匹兹堡地区的一个多模态网络上进行了数值实验，并对需求和管理策略进行了敏感性分析。我们发现，包括乘客总需求、停车价格、交通费用和拼车阻抗在内的许多因素都可以有效地影响系统性能和个人用户成本。在加利福尼亚州Fresno的大型多模态网络上的实验也表明，我们的模型和求解算法具有令人满意的收敛性能和计算效率。
Travel behavior and travel cost in modern urban transportation systems are impacted by many aspects including heterogeneous traffic (private cars, freight trucks, buses, etc.) on roads, parking availability near destinations, and travel modes available in the network, such as solo-driving, carpooling, ride-hailing, public transit, and park-and-ride. Managing such a complex multi-modal system requires a holistic modeling framework of transportation network flow in terms of both passenger flow and vehicular flow. In this paper, we formulate and solve for spatio-temporal passenger and vehicular flows in a general multi-modal network explicitly considering multi-class vehicles, parking facilities, and various travel modes. Vehicular flows, namely cars, trucks, and buses, are integrated into a holistic dynamic network loading (DNL) model. Travel behavior of passenger demand on modes and routes choices is encapsulated by a multi-layer nested logit model. We formulate the multi-modal dynamic user equilibrium (MMDUE) that can be cast into a Variational Inequality (VI) problem. A simple flow solution performed at the origin-destination level and based on the gradient projection is derived from the KKT conditions and shown to efficiently solve for the VI problem on large-scale networks. Numerical experiments are conducted on a multi-modal network in the Pittsburgh region along with sensitivity analysis with respect to demand and management strategies. We show that many factors including the total passenger demand, parking prices, transit fare, and ride-sharing impedance can effectively impact the system performance and individual user costs. Experiments on a large-scale multi-modal network in Fresno, California also show our model and solution algorithm have satisfactory convergence performance and computational efficiency.