Ridesharing services play an essential role in modern transportation, which significantly reduces traffic congestion and exhaust pollution. In the ridesharing problem, improving the sharing rate between riders can not only save the travel cost of drivers but also utilize vehicle resources more efficiently. The existing online-based and batch-based methods for the ridesharing problem lack the analysis of the sharing relationship among riders, leading to a compromise between efficiency and accuracy. In addition, the graph is a powerful tool to analyze the structure information between nodes. Therefore, in this paper, we propose a framework, namely StructRide, to utilize the structure information to improve the results for ridesharing problems. Specifically, we extract the sharing relationships between riders to construct a shareability graph. Then, we define a novel measurement shareability loss for vehicles to select groups of requests such that the unselected requests still have high probabilities of sharing. Our SARD algorithm can efficiently solve dynamic ridesharing problems to achieve dramatically improved results. Through extensive experiments, we demonstrate the efficiency and effectiveness of our SARD algorithm on two real datasets. Our SARD can run up to 72.68 times faster and serve up to 50% more requests than the state-of-the-art algorithms.

翻译：拼车服务在现代交通中扮演着至关重要的角色，它能显著缓解交通拥堵并减少尾气污染。在拼车问题中，提高乘客间的共享率不仅能节省司机的出行成本，还能更高效地利用车辆资源。现有针对拼车问题的在线式与批处理方法缺乏对乘客间共享关系的分析，导致效率与准确性难以兼顾。此外，图是分析节点间结构信息的强大工具。为此，本文提出一种名为StructRide的框架，利用结构信息以改进拼车问题的求解结果。具体而言，我们提取乘客间的共享关系以构建共享图。随后，我们为车辆定义了一种新颖的共享损失度量，用以选择请求组，使得未被选中的请求仍具有较高的共享概率。我们提出的SARD算法能够高效求解动态拼车问题，并获得显著提升的结果。通过大量实验，我们在两个真实数据集上验证了SARD算法的高效性与有效性。相较于最先进的算法，我们的SARD算法运行速度最高可提升72.68倍，并能多服务高达50%的请求。