School of Mathematics and Statistics - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemModels and algorithms for drayage routing problems involving rich characterizations of fleets( 2023-08)Drayage is the trucking of containers over short distances, which connects facilities such as intermodal terminals, distribution centers, storage yards, and locations of shippers and consignees. Drayage is pivotal to intermodal transportation, and although it operates over short distances, it represents an important part of the total intermodal costs. The high economic importance of drayage has motivated an increased interest in the topic, both from academia and logistics practitioners. The planning of drayage movements involves a number of logistic decisions. The academic literature has focused on issues such as empty container repositioning and deciding on whether to separate trucks from containers during packing/unpacking operations. However, questions such as the characterization of truck and container fleets have often been overlooked. With this study, we aim to fill a research gap in this regard. In this study, we tackle drayage routing problems where the characterization of the fleet is essential. The problems we tackle allow for the consideration of containers of any size and cargo category and trucks of different types, with load configuration and compatibility constraints defining which and how many containers a truck can carry. We consider both standard trucks and longer combination vehicles (LCVs). We propose a state transition logic that defines load configuration and compatibility constraints. Based on this logic, we propose two mixed-integer programming (MIP) formulations — suitable for instances involving standard trucks — and an adaptive large neighborhood search (ALNS) heuristic — suitable for instances involving both standard trucks and LCVs. The first MIP formulation is compact and can be solved directly with black-box MIP solvers. The second MIP formulation combines fragments of routes across a space-time network and is solved with a tailored branch-and-cut algorithm. We assess the performances of these formulations through computational experiments on instances with different numbers of requests, geographical distributions of locations, time-window lengths, and fleet compositions (considering only standard trucks). We discuss how these characteristics impact the performances of the formulations. We also show how our formulations can be used to derive managerial insights into different scenarios of drayage routing involving standard trucks. The proposed ALNS heuristic features a novel acceptance criterion and problem-specific search operators that consider issues such as truck/container compatibility and load configurations specific to each truck. We assess this heuristic through computational experiments on instances similar to those used to assess the MIP formulations (in this case, we consider instances involving both standard trucks and LCVs). The obtained results show that the proposed ALNS heuristic can consistently find high-quality solutions with a manageable computational effort. We also use this heuristic to draw managerial insights into the benefits of LCVs over standard trucks.