We address a novel ride‐sharing model aiming to improve the quality of carpool services, by means of incorporation of passengers' efforts. In densely populated areas, locating and picking up passengers are often time costly. These phenomena, which are rarely considered in literature, will hurdle the service efficiency of ride‐sharing businesses by a significant amount, and contribute to the performance gap between theoretical estimation and practical performance. We notice that major platforms, such as DiDi, Uber, and Lyft are all starting innovative ways to encourage passengers to take efforts to walk to the assigned stations. We present a static optimization model for the planning stage of the service and establish an estimation of the carpooling rate for each station together with robustness consideration. In addition, we provide simple and efficient optimization algorithms with theoretical performance guarantee. At the same time, we conduct simulation studies based on both random generated data and open accessible data from DiDi. By applying our model and algorithm to the data of DiDi, we can achieve carpooling rates of 90% while passengers are located within an average radius of 250 m spending an extra waiting time of less than 3 min for their shared rides during rush hour.

中文翻译：

---

结合乘客努力的新乘车共享模型

我们提出了一种新颖的乘车共享模型，旨在通过整合乘客的努力来提高拼车服务的质量。在人口稠密的地区，定位和接送乘客通常会花费大量时间。这些现象在文献中很少被考虑，这将大大削弱乘车共享企业的服务效率，并造成理论估计与实际绩效之间的绩效差距。我们注意到，诸如DiDi，Uber和Lyft之类的主要平台都在开始创新的方式，以鼓励乘客努力前往指定的车站。我们为服务的规划阶段提供了一个静态优化模型，并为每个站点建立了拼车率的估计值，并考虑了健壮性。此外，我们提供具有理论性能保证的简单有效的优化算法。同时，我们基于随机生成的数据和来自DiDi的开放式可访问数据进行仿真研究。通过将我们的模型和算法应用于DiDi的数据，当乘客位于250 m的平均半径内时，在高峰时段共享乘车的额外等待时间少于3分钟，我们可以实现90％的拼车率。