Optimization Methods Applied in a Class of Vehicle Routing and Scheduling Problems

Κωνσταντακόπουλος, Γρηγόριος; Konstantakopoulos, Grigorios

dc.contributor.author	Κωνσταντακόπουλος, Γρηγόριος	el
dc.contributor.author	Konstantakopoulos, Grigorios	en
dc.date.accessioned	2022-11-23T08:41:47Z
dc.date.available	2022-11-23T08:41:47Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/56222
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.23920
dc.rights	Default License
dc.subject	Vehicle Routing Problem	en
dc.subject	Optimization Algorithms	en
dc.subject	Software	en
dc.subject	Scheduling
dc.subject	Logistics
dc.title	Optimization Methods Applied in a Class of Vehicle Routing and Scheduling Problems	en
dc.title	Μέθοδοι Βελτιστοποίησης για Ένα Σύνολο Προβλημάτων Δρομολόγησης Οχημάτων και Προγραμματισμού Παραδόσεων	el
dc.contributor.department	Βιομηχανικής Διοίκησης και Επιχειρησιακής Έρευνας	el
heal.type	doctoralThesis
heal.classification	optimization algorithms	el
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2022-07-05
heal.abstract	The present thesis studies and addresses a class of vehicle routing and scheduling problems while also developing powerful evolutionary and local search metaheuristics that aim to optimize the distribution process in urban areas by proposing a plan of routes and deliveries. The fact that route planning is affected by many variables and constraints that have emerged by the nature of logistics companies, their external environment, and their customers' needs and requirements has led to an increased interest both scientifically and commercially. The current thesis, aiming to bring together research and industry, also studies and proposes the integration of the developed algorithms into a cloud routing system, offering highly efficient route plans. Moreover, the different variables and constraints of the Vehicle Routing Problem (VRP) correspond to different variants of the problem. In the current thesis, the variants considered have arisen through an analytical literature review and the cooperation with industrial and software development companies operating in this field. Specifically, the most commonly addressed, by researchers and practitioners, VRP variants and those with the potential to contribute financially and environmentally to logistics companies in the future are selected. This methodology ensures that both fields (research and industry) will benefit and be interested in, as they are inextricably linked to each other. The variants of the problem that are selected and studied are: • the Vehicle Routing Problem with Time Windows (VRPTW), • the Vehicle Routing Problem with Simultaneous Pickups and Deliveries (VRPSPD), • the Heterogeneous Fleet Vehicle Routing Problem (HVRP), and • the Collaborative VRP. The VRP variants under study belonging to NP-hard problems are highly complex, primarily when addressed simultaneously. In this thesis, one of the goals is to address all the above-mentioned variants either solely or combined in order to cover most needs and requirements of logistics companies operating on freight distribution field. Each of the above VRP variants corresponds to specific characteristics, variables and restrictions. Specifically, the VRPTW considers the time slots that customers indicate in order to be served by logistics companies. The VRPSPD focuses, as its name indicates, on simultaneous pickups and deliveries, while the HVRP on the heteroginity of vehicles that logistics companies own, in terms of capacity, fixed and variable costs. Finally, the Collaborative VRP focuses on cases that logistics companies that locate and operate at the same areas collaborate in the distribution process. All these VRP variants form the main requirements and challenges that logistics companies face on freight distribution, and by addressing them great benefits can emerge. Combining multiple VRP variants is complex, along with the need of companies to acquire efficient and cost-saving solutions have led to the need for optimization algorithms. Initially, exact algorithms were proposed by researchers. Their advantage, which is optimally solving the VRP, is accompanied by a significant constrain; the limited number of customers that can be handled and the increased computation time needed, especially when customers exceed 100. This limitation is restrictive, and therefore these algorithms are not integrated into software packages. The disadvantages of exact algorithms resulted in researchers proposing heuristics algorithms that offer good quality solutions that are not optimal in most cases in limited computation time. Finally, researchers developed metaheuristic algorithms that, in most cases, offer the best trade-off between efficiency of solutions and computation time. In the current thesis, multiple metaheuristic algorithms are studied, developed, and proposed to solve the VRP variants that are mentioned above and correspond to those that logistics companies most commonly address on freight distribution. An Evolutionary Algorithm (EA) and two Local Search (LS) metaheuristics are developed. These algorithms address either separately or simultaneously some of the proposed VRP variants. More precisely, the developed EA addresses the Vehicle Routing Problem with Simultaneous Pickups and Deliveries and with Time Windows (VRPSPDTW). The algorithm is also applied in cases with electric vehicles with a limited driving range, affecting distribution. The novelty in this solution method is related to the applied crossover operator and the fact that the problem is addressed as multiobjective for the first time. In multiobjective optimization, two or more conflicting objectives exist and need to be optimized, which in the current research are (i) the number of vehicles needed for executing the routes, and (ii) the total traveled distance. Additionally, the first LS algorithm, a Large Neighborhood Search (LNS) metaheuristic, addresses the VRPTW as a multiobjective. This is the first time a multiobjective LNS algorithm is proposed in the specific problem, offering high-quality results. Finally, the second developed LS metaheuristic is an Adaptive Large Neighborhood Search (ALNS) metaheuristic that addresses the Heterogeneous Fleet Vehicle Routing Problem with Simultaneous Pickups and Deliveries and with Time Windows (HVRPSPDTW), which means that the three VRP variants are considered simultaneously. The ALNS algorithm is also applied, after some modifications, in the Collaborative VRP to estimate the economic and environmental benefits from the cooperation of distribution companies. In the context of studying VRP variants and addressing them through optimization algorithms, their mathematical model was formulated, as it is a vital part of efficiently addressing the problem and setting the right variables and constraints. Finally, the three algorithms have been developed in the Python programming language as libraries containing the necessary functions for solving the VRP. The algorithms are tested in multiple datasets of the literature to ensure their efficiency and ability to be integrated into a cloud routing system. The proposed system is developed in cooperation with a software development company that is responsible for the user interface part, as well as the connection of the cloud system with its individual parts of the system (such as online maps and the database). Additionally, the requirements of the system regarding the master data, as well as its functionalities have been defined jointly. On the other hand, the core of the system, which is the algorithms that are implemented, have been developed exclusively by me, and configured to serve the need of the software development company. Since all of the proposed algorithms offer high-quality results, they are all stored in the software development company’s server, so that to be accessible by the cloud routing system through the appropriate requests. While the system has the ability to access any of the proposed algorithms, in its current form, the system receives the plan of routes that is extracted by the ALNS algorithm, covering multiple freight distribution cases. The system that handles and addresses multiple variables and constraints needs to be fed with the correct data to offer efficient routes and deliveries. Therefore, it exploits advanced technologies that can ensure the quality and accuracy of data. Online maps are exploited concerning the geocoding procedure and the data related to the distance and the travel time between order points. The benefits of using such a routing system provided through the cloud, and thus no installation of the system is required on each company's computers, are many and are not limited only to the financial aspect. Besides eliminating installation and maintenance costs, the users can extract efficient and high accuracy plans that are visualized on online maps, modify the proposed routes and schedules, and execute rerouting strategies after the distribution process has started. In addition to these functionalities, when fed with real-time data from online maps, the system can handle the dynamic nature of the problem and address unexpected events that are occurred after the plan of routes have begun. Specifically, data may change during the route execution, affecting the distribution process. All the technologies mentioned above and system functionalities can significantly enhance logistics companies' efforts and decision-makers before and after the distribution process started. Concluding, the research focuses on multiple aspects, starting with the study of the above-mentioned VRP variants which are considered either solely or combined, along with their mathematical formulation. In addition, the presented thesis proposes and develops optimization algorithms that address these distribution cases and are tested in both datasets of the literature and in real-life cases in order to evaluate the ability of the algorithms to be implemented in a routing system. The last part of the research concerns the ability of algorithms to be suitable for integration into a cloud routing system, as well as their connection with technologies that are necessary for system development.	en
heal.advisorName	Τατσιόπουλος, Ηλιάς	el
heal.advisorName	Tatsiopoulos, ilias	en
heal.committeeMemberName	Πόνης, Σταύρος
heal.committeeMemberName	Παναγιώτου, Νικόλαος	el
heal.committeeMemberName	Αραβώσης, Κωνσταντίνος	el
heal.committeeMemberName	Τόλης, Αθανάσιος
heal.committeeMemberName	Βαρβαρίγου, Θεοδώρα	el
heal.committeeMemberName	Ζεϊμπέκης, Βασίλειος	el
heal.committeeMemberName	Τατσιόπουλος, Ηλιάς	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Αγρονόμων και Τοπογράφων Μηχανικών	el
heal.academicPublisherID	ntua
heal.numberOfPages	302
heal.fullTextAvailability	false