| Author: | Xin, Xu |
| Title: | Data-driven distributionally robust optimization for resilient maritime logistics systems under uncertainty |
| Advisors: | Wang, Shuaian Hans (LMS) |
| Degree: | Ph.D. |
| Year: | 2025 |
| Department: | Department of Logistics and Maritime Studies |
| Pages: | xv, 215 pages : color illustrations |
| Language: | English |
| Abstract: | The maritime industry is intricately linked to global political and economic conditions, making its inherent uncertainties a central concern for both academia and industry. Recent years have seen a surge in unforeseen events (e.g., the COVID-19 pandemic), frequently subjecting stakeholders to volatile container transportation demand, port congestion, and shipment delays. In this context, the development of a maritime logistics system capable of maintaining an acceptable level of maritime service continuity in the face of uncertainty—or, more specifically, the establishment of a resilient maritime logistics system—has become a critical measure for enhancing global supply chain resilience. Motivated by the above background, this thesis addresses the crucial challenge of improving the resilience of maritime logistics systems under two primary uncertain factors: (1) uncertain vessel delay times and (2) uncertain container transportation demand. In order to comprehensively enhance the resilience of various elements within the maritime logistics system (i.e., ports and shipping routes), this thesis investigates three distinct problems. The first problem is berth allocation (landside problem), with a focus on improving port operation resilience. The second problem is liner fleet deployment (seaside problem), with a focus on improving shipping route resilience. The third problem is the joint optimization of liner fleet deployment and slot allocation (seaside problem), with a focus on improving shipping route resilience. Through three interconnected studies, detailed in Chapters 3 to 5, we develop novel distributionally robust optimization (DRO) frameworks that effectively mitigate the limitations of traditional stochastic programming (SP) and robust optimization (RO) modeling approaches. Furthermore, we design customized algorithms and acceleration strategies to efficiently solve large-scale problems. First, we present a distributionally robust chance-constrained (DRCC) model with a type-1 Wasserstein ambiguity set to address the landside berth allocation problem under vessel arrival time uncertainty in Chapter 3. This model overcomes the conservatism of robust optimization and eliminates the distributional assumption requirement of stochastic programming. Numerical experiments validate that introducing acceleration strategies, i.e., sample average approximation constraints and coefficient strengthening techniques, significantly reduce program running time. Extensive sensitivity analyses reveal how pre-defined service levels in the chance constraint and ambiguity set radii impact operational decisions. We also compare the effects of employing a joint chance constraint versus several independent chance constraints on the solution. The out-of-sample analysis results demonstrate the strong out-of-sample performance of our DRCC model while meeting port service requirements. Then, for the seaside liner fleet deployment problem under container transportation demand uncertainty, we propose a data-driven DRCC model also adopting a type-1 Wasserstein ambiguity set in Chapter 4. This approach minimizes the total operational costs for liner companies while ensuring that the capacity deployed to each route meets transportation demand through a joint chance constraint. We reformulate the DRCC model into a mixed integer programming model, accelerating its solution through introducing sample average approximation constraints and several valid inequalities. Sensitivity analysis on the ambiguity set radius and the pre-defined service level in the joint chance constraint is conducted. Moreover, we compare the optimal solutions obtained by stochastic programming and robust optimization with solution obtained by the DRCC model. The out-of-sample test results confirm that liner companies can achieve a higher rate of transportation demand fulfillment based on the DRCC model, and our DRCC model outperforms robust optimization model (by avoiding excessive conservatism) and stochastic programming model (by preventing optimization bias). Finally, to jointly optimize liner fleet deployment and slot allocation under container transportation demand uncertainty, we establish a two-stage distributionally robust model with two ellipsoidal ambiguity sets in Chapter 5. Similar to the model in Chapter 4, this distribution-free framework also incorporates a joint chance constraint to ensure that the weekly container transportation demands of contract shippers are met to a given service level. We reformulate the original model into a second-order cone model (SOCM), which is efficiently solved via a customized outer approximation algorithm. Sensitivity analyses are performed on several key parameters, including the ambiguity set radius, the mean and covariance of contract demand, and liner capacity. Out-of-sample comparisons empirically verify the superiority of our DRO model. Collectively, these studies contribute to the field of maritime logistics system by providing cutting-edge modelling tools and DRO frameworks to address critical operational challenges under uncertainty. The developed models and algorithms offer practical solutions for enhancing the resilience of maritime logistics systems, directly supporting stakeholders in navigating volatile market conditions and improving the robustness of global maritime supply chains. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14161