Efficient active-passive vehicle coordination in multimodal transportation networks

This dataset accompanies a study submitted to "Transportation Research Part B: Methodological", focusing on a modular, demand-responsive transport system. The system uses passive, modular containers, or “pods,” that are transported across multiple vehicle types like trucks, trains, and boats to fulfill door-to-door customer requests. The study's main objective is to improve customer satisfaction and reduce operator costs, thus providing a more flexible alternative to traditional public transportation systems. The unique challenge lies in synchronizing the movements of both the pods and the vehicles they rely on, which creates interdependencies between routes. To address this, a mixed-integer linear program (MILP) and an adaptive large neighborhood search (ALNS) heuristic were developed, alongside a clustering approach to accelerate solution times for real-world applications. This dataset validates the performance of these optimization methods, providing a practical framework for testing advanced multimodal routing solutions.

Since no standard benchmark data existed for this novel system, the dataset consists of three custom instance sets - A, B, and C - with increasing levels of complexity and sizes, containing 30 instances each. Set A includes scenarios with 4 customer requests, 2 pods, 2 bases, and 6 hubs, representing smaller, simpler problems. Set B scales up to 10 requests, 6 pods, 4 bases, and 8 hubs, while Set C represents the most complex instances with 20 requests, 10 pods, 5 bases, and 10 hubs. Each instance includes randomly generated hubs in a two-dimensional space, with minimum pairwise distances maintained for spatial realism. The transportation network is constructed by assigning edges randomly within each modality (e.g., road, rail) while ensuring full connectivity. Additionally, each base and pod is assigned a capacity, and time windows for customer requests vary, with each subsequent set having a longer time window range. Travel times are computed as the Euclidean distance between hubs divided by a constant speed factor, with adjustments for loading and boarding times to reflect real-world operational delays.

本数据集伴随一项提交至《交通运输研究B：方法论》的研究，该研究聚焦于一种模块化、响应需求的运输系统。该系统采用被动式、模块化容器，或称之为“舱”，这些舱可通过多种车型，如卡车、火车和船只进行运输，以满足门到门的客户需求。研究的主要目标是提升客户满意度并降低运营商成本，从而为传统公共交通系统提供一种更为灵活的替代方案。独特的挑战在于同步舱体及其依赖的运输工具的运动，这导致了路线间的相互依赖。为应对此挑战，研究团队开发了一种混合整数线性规划（MILP）、自适应大邻域搜索（ALNS）启发式算法以及一种聚类方法，以加速实际应用中的解决方案时间。该数据集验证了这些优化方法的有效性，为高级多模式路由解决方案的测试提供了一个实用的框架。鉴于该新颖系统尚无标准基准数据，数据集包含三个定制实例集——A、B和C，其复杂性和规模依次递增，每个集合包含30个实例。集合A包括具有4个客户请求、2个舱、2个基地和6个枢纽的情景，代表较小、较简单的问题。集合B扩展至10个请求、6个舱、4个基地和8个枢纽，而集合C则代表了最复杂的实例，包含20个请求、10个舱、5个基地和10个枢纽。每个实例都包含在二维空间中随机生成的枢纽，并保持成对的最小距离，以维持空间真实性。运输网络通过在每个模式（例如，道路、铁路）内随机分配边来构建，同时确保完全连通。此外，每个基地和舱都被分配了一个容量，客户请求的时间窗口也各不相同，后续集合的时间窗口范围更长。旅行时间通过将枢纽之间的欧几里得距离除以一个恒定速度因子来计算，并对装载和登乘时间进行调整，以反映现实世界的运营延迟。