Routing optimization for emerging transportation systems

Abstract

This thesis investigates two routing optimization problems in two emerging transportation systems correspondingly, where the first one relates to unmanned surface vehicles and the second one relates to floating targets. The first problem is formalized as a clustered coverage orienteering problem, which is a generalization of the classical orienteering problem. The problem is widely motivated by the emerging unmanned techniques (e.g., unmanned surface vehicles and drones) applied to environmental monitoring. Specifically, in this study, the unmanned surface vehicles are used to monitor reservoir water quality by collecting samples. In this problem, the water sampling sites (i.e., the nodes) are grouped into clusters, and a minimum number of nodes must be visited in each cluster. With each node representing a certain coverage area of the water, the objective is to monitor as much as possible the total coverage area in one tour of the unmanned surface vehicle, considering that overlapping areas provide no additional information. An integer programming model is first formulated through a linearization procedure that captures the overlapping feature. A two-stage exact algorithm is proposed to obtain an optimal solution to the problem. The efficiency and effectiveness of the two-stage exact algorithm are demonstrated through experiments on randomly generated instances. The second problem explores routing optimization with floating targets. Several modern transportation and logistics systems make use of floating targets, which allow vehicles to choose where to visit customers rather than stopping in pre-determined locations. Floating targets introduce an additional degree of freedom but complicate routing optimization. This study tackles the vehicle routing problem with floating targets and the dial-a-ride problem with floating targets. These problems can be formulated as mixed-integer second-order cone programs (with the Euclidean distance) or mixed-integer linear programs (with the Manhattan distance), but are too complex to be solved efficiently in large-scale instances using off-the-shelf methods. To address these challenges, this study develops an exact decomposition algorithm. First, we leverage the geometric structure of operations to optimize the pickup location of a single stop. Second, we develop an iterative procedure to decompose a multi-stop problem into a series of one-stop problems, until convergence to the globally optimal solution. Third, we embed this procedure into dynamic programming algorithms for solving the vehicle routing problem with floating targets and the dial-a-ride problem with floating targets. Computational results show that our solution approach results in superior solutions and shorter computational times than state-of-the-art benchmarks.