城市的发展依赖于道路交通系统的高效运转。然而，城市化、机动化程度的不断加深使得道路交通供需关系日趋紧张，导致了严重的交通外部性影响。随着控制技术与经济模式的发展，自动驾驶与共享出行将成为未来城市交通的两大主题，为解决城市道路交通系统外部性问题提供新的途径。本文致力于研究共享自动驾驶出行的运营优化策略，开展共享自动驾驶车辆（Shared Autonomous Vehicle, SAV）在服务状态下的调度优化以及在非服务状态下的停车优化，进而量化该出行模式给城市道路交通系统带来的潜在影响。首先，本文提出基于车牌识别数据的车辆轨迹重构方法，将路侧节点检测数据转换为含有时空信息的轨迹数据，进而在城市规模车牌识别数据样本集中提取机动化出行需求信息，为共享自动驾驶出行运营优化建模提供数据支撑。其次，本文探索了面向人工驾驶车辆的共享出行模式及其交通影响。基于改进的最大公共子序列算法分析轨迹的时空相似性，搜索潜在的合乘匹配对。构建大规模整数规划模型提取最优的合乘匹配组合，使得路网流量缩减达到最大化。本部分为探究共享自动驾驶出行提供方法基础与前景描述。进一步，本文讨论面向SAV的共享出行，探究SAV在服务状态下的运营优化策略及其交通影响。本文关注一种未来潜在的出行场景，即人们不再购买和保有车辆，而是使用SAV车队提供的车辆共享与行程共享服务。本文设计了基于旅行时间约束的行程共享匹配准则，搜索潜在的行程共享匹配对，通过大规模整数规划求解最优的行程共享匹配组合；构建车辆指派模型，并将其转化为网络规划中的最小路径覆盖问题，使用图与网络算法进行求解，实现以最小的车队规模满足所有的机动化出行需求。本文进一步开展多场景SAV共享出行效益分析。具体而言，本文探究在不同公众参与度场景、不同城市规模场景、以及考虑公众实际参与意愿场景下的SAV共享出行，揭示SAV推广过程中的潜在效益与问题，说明该出行模式的适用性。最后，本文建立了面向SAV的停车设施规划方法，探究SAV在非服务状态下的运营优化策略。构建两阶段随机优化模型，以最低的社会成本规划停车设施满足SAV临时停车需求。第一阶段决策生成停车设施的位置及容量；第二阶段生成补偿决策，通过SAV停车调度提高第一阶段规划方案的适应能力。本文将所提出的停车规划模型应用于城市案例，讨论该模型在处理需求不确定性方面的优越性。

Urban development depends on the efficient operation of the road transportation system. However, the deepening of urbanization and motorization has made the relationship between demand and supply in road transportation increasingly tense, which further causes extreme traffic externalities. With the development of technology and economy, autonomous driving and shared mobility will become the two major directions of the future urban transportation system, leading to a promising solution to cope with the externalities of urban transportation.This study is dedicated to researching the operational optimization of mobility with shared autonomous vehicles （SAVs）. This study conducts the assignment optimization of SAVs in the service state and the parking optimization of SAVs in the non-service state. Further, this study quantifies the potential impact of mobility with SAVs on the urban road transportation system.Firstly, this study proposes a vehicle trajectory reconstruction method based on license plate recognition data, converting roadside node-based detection data into trajectory data that contains spatiotemporal information. Furthermore, the trajectory reconstruction method is applied to urban-scale license plate recognition data to extract motorized travel demand information, which provides data support for the operational optimization modeling of mobility with SAVs.Secondly, this study investigates the operation and impact of carpooling with human-driven vehicles. An updated longest common subsequence algorithm is used to evaluate the spatiotemporal similarity of trajectories and retrieve all potential carpooling matching pairs. An integer programming model is constructed to find the optimal pairwise matching, which leads to the maximum reduction in traffic volume. This part provides methodology foundation and prospect description for the investigation on the mobility with SAVs.Furthermore, this study investigates the mobility with SAVs, demonstrating the operation strategy and traffic impact of SAVs in service state. This study focuses on a future mobility scenario, where the individuals no longer purchase or own vehicles, but use the ride-sharing and car-sharing services provided by the SAV fleet. This study designs a travel-time-constrained ride-sharing matching strategy, retrieving all the potential ride-sharing matching pairs. A large-scale integer programming model is formulated to search for the optimal pairwise ride-sharing matching. A vehicle assignment model is constructed to search for the minimum fleet to fulfill the travel demand. To make the vehicle assignment model tractable, the original model is transformed into a minimum path cover problem in the network programming, which can be exactly solved by a graph-based algorithm. This study further conducts multi-scenario analysis to demonstrate the feasibility and impact of SAVs. Specifically, this study investigates mobility with SAVs under cases with different shared mobility participation levels, in case cities with different sizes, and in a scenario considering actual public participation willingness. This part helps reveal potential benefits and possible problems during the promotion of SAVs. Finally, this study carried out parking planning towards SAVs, demonstrating the operational optimization of SAVs in non-service state. Two-stage stochastic parking facility planning models are constructed to fulfill the temporary parking demand of SAVs with the minimum social cost. The first-stage decision aims to determine the location and capacity of the parking facilities. The second stage generates recourse decisions, which return the parking relocation assignments of SAVs to make the first-stage planning scheme adapt to various uncertain demand scenarios. This model is applied to a real-world case to show its effectiveness.