This study investigates a flexible one-to-two matching problem in ride-hailing systems. In the ride-hailing platform, dedicated drivers are assigned to serve a single request (one-to-one matching) or two requests (one-to-two matching). We aim to examine the en-route matching problem, i.e., an in-transit driver who is currently serving riders may be informed to pick up co-riders traveling in the same direction. We consider boundedly rational users who accept ride-shares at reasonable travel costs. To encourage riders to join one-to-two matching, we propose an incentive scheme under which riders who experience detour time in one-to-two matching will be charged at a discounted rate. To make en-route decisions quickly, we propose a rolling horizon approach. The flexible one-to-two matching problem is formulated as a multi-stage integer program with the objective of maximizing the system-wide profit. We derive the upper bound on detour time for riders, based on which we develop an efficient algorithm to determine cost-feasible ride-shares. We also develop a neighborhood searching method to identify the locations that are close to a driver. Exploiting the historical trip data of New York taxis, we investigate the matching patterns and provide strategies to benefit the users and the platform when the ride-hailing program is introduced into the existing road network. The numerical experiments reveal that our proposed incentive scheme for one-to-two matching problems can increase both ridership and profit. Furthermore, the productivity and the compensation of dedicated drivers are also improved.

这项研究调查了叫车系统中灵活的一对二匹配问题。在乘车平台中，分配了专用的驱动程序来满足一个请求（一对一匹配）或两个请求（一对二匹配）。我们旨在研究路途匹配问题，即，可能会告知当前正在为骑手提供服务的在途驾驶员提车沿相同方向行驶的搭便车。我们考虑以合理的旅行费用接受拼车的有限理性用户。为了鼓励骑手加入一对一比赛，我们提出了一项激励计划，根据该计划，将以折扣价向在一对一比赛中经历de回时间的骑手收费。为了快速做出在途决策，我们提出了一种滚动式方法。灵活的一对二匹配问题被公式化为一个多级整数程序，目的是使系统范围内的利润最大化。我们得出了车手绕行时间的上限，在此基础上，我们开发了一种有效的算法来确定成本可行的乘车份额。我们还开发了一种邻域搜索方法，以识别靠近驾驶员的位置。利用纽约出租车的历史旅行数据，我们研究了匹配模式，并提供了在将打车程序引入现有道路网络时使用户和平台受益的策略。数值实验表明，我们提出的针对一到两个匹配问题的激励方案可以增加乘车人数和利润。此外，还提高了生产率和专用驱动器的补偿。