Mobility on demand has been gaining more attention from the research community as a way to offer smart and efficient transportation services to people. Despite the advancements in vehicular technologies, vehicle breakdown (VB) remains one of the major contributors to the disruption of fleet operations, which may inflict large recovery costs and damage the service provider?s reputation. However, modeling such dynamic events for dial-a-ride (DAR) systems has not been addressed until now; this is the research gap we attempt to address in this article. The following are the contributions of our work: 1) the formulation of a disruptive DAR problem (DDARP) with VB (DDARP-VB), 2) a GPU-based adaptive large neighborhood search (G-ALNS) algorithm to solve the DDARP-VB, and 3) a fleet size minimization (FSM) strategy that leads to reduced operational costs under disruptions. From our simulations, the proposed G-ALNS algorithm performs up to 52 times faster than its CPU counterpart and produces better solutions when compared to the existing GPU approach for DARPs. Overall, our FSM strategy leads to a 59% reduction in fleet idle time under normal operations and a 15% reduction in operational cost under disruption.

中文翻译：

---

Dial-A-Ride 系统的中断管理

作为一种为人们提供智能高效交通服务的方式，按需出行越来越受到研究界的关注。尽管车辆技术取得了进步，但车辆故障 (VB) 仍然是造成车队运营中断的主要因素之一，这可能会造成巨额恢复成本并损害服务提供商的声誉。然而，直到现在还没有解决为拨号搭车 (DAR) 系统建模此类动态事件。这是我们试图在本文中解决的研究空白。以下是我们工作的贡献：1) 使用 VB (DDARP-VB) 制定破坏性 DAR 问题 (DDARP)，2) 基于 GPU 的自适应大邻域搜索 (G-ALNS) 算法来解决 DDARP- VB，3) 机队规模最小化 (FSM) 策略，可在中断情况下降低运营成本。根据我们的模拟，与现有的 DARP 的 GPU 方法相比，所提出的 G-ALNS 算法的执行速度比其 CPU 对应算法快 52 倍，并产生更好的解决方案。总体而言，我们的 FSM 战略可将正常运营下的机队闲置时间减少 59%，并将运营成本降低 15%。