Ride-sharing services, that have been growing in recent years with the start of network service companies, will be further enhanced by the recently emerging trend of applications for autonomous vehicles for future traveler mobility. One fundamental question that transportation managers should address is how to capture the endogenous traffic patterns involving the new and uncertain elements facing future transportation planning and management. By concentrating on one ideal system optimal (SO) scenario, in which (i) all vehicles are autonomous, or can be centrally guided and (ii) all passengers’ pickup/drop-off trip requests can be given at the beginning, this paper aims to integrate travel demand, vehicle supply, and limited infrastructure. Available ride-shared and autonomous vehicles, from different (real/virtual) depots, can be optimally assigned to satisfy passengers’ trip requests, while considering the endogenous congestion in capacitated networks. A number of decomposition approaches are adopted in this research. Focusing on this primal problem, we propose an arc-based vehicle-based integer linear programming model in space-time-state (STS) networks, which is solved by Dantzig-Wolfe decomposition. From the perspective of dynamic traffic assignment, a space-time-state (STS) path-based flow-based linear programming model is also provided as an approximation according to the mapping information between vehicle and passenger, and between a vehicle and the space-time arc in each STS path in our priori-generated column pool. Finally, numerical experiments are performed to demonstrate our decomposition approaches and their computation efficiency. From our preliminary experiments, we have a few interesting observations: (i) without considering road congestion, the network performance/efficiency could be overestimated; (ii) passengers’ required pickup and drop-off time windows could be a buffer to mitigate road congestion, without impacting system performance; (iii) the ride-sharing service could reduce the total transportation system cost under centralized control.

随着网络服务公司的成立，近几年来共享乘车服务不断增长，而自动驾驶汽车为未来旅行者出行带来的最新趋势将进一步增强这种共享服务。运输管理人员应解决的一个基本问题是如何捕获涉及未来运输规划和管理面临的新的不确定因素的内生交通模式。通过专注于一种理想的系统最佳（SO）方案，在该方案中（i）所有车辆都是自主的，或者可以集中引导，并且（ii）可以在一开始就给出所有乘客的上落客请求旨在整合旅行需求，车辆供应和有限的基础设施。来自不同（真实/虚拟）仓库的可用乘车共享和自动驾驶汽车，可以在考虑到容量网络中的内源性拥塞的同时，为满足旅客的出行需求而进行最佳分配。本研究采用了许多分解方法。针对这个主要问题，我们提出了一种基于时空状态（STS）网络的基于弧的基于车辆的整数线性规划模型，该模型通过Dantzig-Wolfe分解得以解决。从动态交通分配的角度来看，还根据车辆和乘客之间以及车辆与空间之间的映射信息，提供了基于空时状态（STS）路径的基于流的线性规划模型作为近似值。先验生成的列池中每个STS路径中的时间弧。最后，通过数值实验证明了我们的分解方法及其计算效率。从我们的初步实验中，我们有一些有趣的发现：（i）如果不考虑道路拥堵，网络的性能/效率可能会被高估；（ii）乘客所需的上下车时间窗口可能是缓解道路拥堵而又不影响系统性能的缓冲；（iii）拼车服务可以降低集中控制下的运输系统总成本。