Ride Shareability in Shanghai

在城市交通拥堵与碳排放压力日益加剧的背景下，共享出行被广泛视为提升出行效率、减少车辆使用与降低环境负担的重要路径。 相比单人出行，拼车出行(ride sharing)通过合并具有相似时空需求的出行请求，有望显著减少车辆行驶里程与能源消耗。

然而，其实际效果并不只取决于技术平台或用户意愿，更深层地受到城市空间结构与交通系统运行机制的制约。围绕这一问题，我们团队开展了一系列 面向“共享出行潜力(Ride-Shareability)”的城市建模研究。

我们以百万级出租车轨迹为基础，将个体出行转化为可计算的“共享网络”，通过优化算法求解全局最优拼车方案，从而刻画城市中出行可被共享的理论上限。

在此基础上，引入多尺度统计模型与可解释机器学习方法，系统分析建成环境（密度、功能混合、中心结构、交通运行等）对拼车潜力的影响及其非线性机制。

同时，将优化后的出行方案嵌入交通排放模型与速度—流量关系模型中，评估共享出行在真实道路网络中的减排效应与空间分布特征。

Amid rising congestion and carbon emissions, shared mobility is widely seen as a key pathway to improve efficiency, reduce vehicle use, and lower environmental impacts. Compared with single-occupancy travel, ride sharing pools trips with similar spatiotemporal demands, offering substantial reductions in vehicle miles traveled and energy use.

However, its effectiveness is fundamentally shaped by urban spatial structure and transport system dynamics, beyond platforms or user preferences. To address this, our team conducts urban modeling research on “ride-shareability.”

Using large-scale taxi trajectories, we construct a shareability network and solve for optimal matching to estimate the upper bound of shareable trips. We then combine multi-scale statistical models and explainable machine learning to examine how built environment factors—density, land-use mix, spatial structure, and traffic conditions—affect shareability and its nonlinear patterns. Finally, we integrate optimized trip configurations with emission and speed–flow models to assess the environmental impacts and their spatial distribution.

我们发现，ride sharing的效率提升具有明显的“结构性门槛”：当出行需求达到一定密度后，共享潜力会快速跃升。 城市中心区及交通枢纽周边具有更高的匹配效率，而适度的交通速度与土地混合度能够进一步强化这一过程。 同时，空间结构并非越分散越优，过度多中心化可能削弱出行共享潜力，从而限制共享效率。

在环境效应方面，大规模拼车在优化情景下可显著降低整体车辆行驶与排放水平，但其收益在空间上呈现出不均衡分布。 研究表明，在全局最优拼车情景下，系统层面的车辆行驶里程与燃料消耗可实现约15%–23%的下降，并伴随显著的碳排放与污染物减排。

更重要的是，共享出行通过减少路网中的车辆数量，还能够带来“速度提升”的间接效应，即交通流密度下降推动道路平均车速上升，从而进一步降低单位里程排放水平。

这些收益在空间上呈现出显著的不均衡分布：减排效果在高流量、重污染的主干道路和中心城区更加集中，体现出出行需求集聚带来的规模效应。

总体而言，我们基于大规模时空数据，整合优化算法、机器学习与交通模拟，揭示了共享出行与城市空间结构、低碳转型之间的内在联系，并为自动驾驶时代的交通规划提供定量分析框架。

We find that ride sharing exhibits clear structural thresholds: once demand reaches a certain density, shareability rises sharply. Central areas and transport hubs show higher matching efficiency, further enhanced by moderate traffic speeds and land-use mix. In contrast, excessive polycentricity may weaken trip overlap and limit sharing efficiency.

Meanwhile, large-scale ride sharing can substantially reduce overall travel and emissions under optimized scenarios. At the system level, vehicle miles traveled and fuel consumption can decline by about 15%–23%, with corresponding reductions in carbon and pollutant emissions.

Moreover, by reducing the number of vehicles on the road, ride sharing generates an indirect “speed effect”: lower traffic density increases average speeds, further reducing emissions per kilometer. These benefits are spatially concentrated, with greater reductions on high-traffic, high-emission corridors and in central areas.

Overall, leveraging large-scale spatiotemporal data, we integrate optimization, machine learning, and transport simulation to reveal the links between shared mobility, urban spatial structure, and low-carbon transition, providing a quantitative framework for future transport planning in the era of autonomous mobility.