Data-driven optimization approaches for dynamic urban logistics operational problems

• Main Author: YANG, Jingfeng
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2023
• Subjects: private domain strategy; private domain degree; sales performance; store grade; brand category; Databases and Information Systems
• Online Access: https://ink.library.smu.edu.sg/etd_coll/537; https://ink.library.smu.edu.sg/context/etd_coll/article/1535/viewcontent/GPIS_AY2018_PhD_Yang_Jingfeng.pdf
• Institution: Singapore Management University

Given the rapid pace of urbanization, there is a pressing need to optimize urban logistics delivery operations for enhanced capacity and efficiency. Over recent decades, a multitude of optimization approaches have been put forth to address urban logistics challenges, encompassing routing and scheduling within both static and dynamic contexts. In light of the rising computational capabilities and the widespread adoption of machine learning in recent times, there is a growing body of research aimed at elucidating the seamless integration of data and machine learning within conventional urban logistics optimization models. Additionally, the ubiquitous utilization of smartphones and internet innovations presents novel research challenges in the realm of urban logistics, notably in the domains of last-mile delivery collaboration and on-demand food delivery services. The necessity of addressing these emerging challenges is what motivates my doctoral research, with a focus on the investigation of data-driven optimization methodologies. This thesis will encompass a comprehensive discussion of my research conducted in three key domains: (1) collaborative urban delivery with alliances; (2) dynamic service area sizing optimization for on-demand food delivery services; and (3) optimization of dynamic matching time intervals for on-demand food delivery services. The specific details are outlined as follows: 1. We study the pickup and delivery problem within a collaborative framework, focusing on multiple small Logistics Service Providers (LSPs) operating in an urban setting. These LSPs establish trusted alliances, enabling the shared execution of delivery tasks. Specifically, we address a prevalent challenge in urban logistics: the daily operational tasks of LSPs involve collecting goods from one location and delivering them to another, with each request featuring a delivery time window. We formulate this problem as a Mixed-Integer Programming (MIP) model. To manage the substantial daily delivery volume effectively, we introduce a two-stage approach. First, we determine the allocation of LSP requests among alliances, followed by vehicle routing optimization within each alliance. In the first stage, we propose machine learning models to learn the values of delivery costs from past delivery data, which serve as a surrogate for deciding how requests are assigned. In the second stage, we introduce a tabu search heuristic. Experimental results on a standard dataset show that our proposed learning-based optimization framework is efficient and effective in outperforming the direct use of tabu search in most instances. Using our approach, we demonstrate that substantial savings in costs and, hence, improvement in sustainability can be achieved when these LSPs form alliances and requests are optimally assigned to these alliances. 2. We investigate the combined demand and supply management for on-demand food delivery services by adjusting the radius of their customer service and driver dispatch areas. For each restaurant, the platform needs to decide the (1) customer service area, i.e., the radius of the area within which the customers can see the restaurant’s information and order food from it; and (2) driver dispatch area, i.e., the radius of the area within which the drivers can see the restaurant’s information and deliver orders from it. Leveraging a real dataset from a food delivery platform, we propose a data-driven optimization framework that combines machine learning methods for order delivery time estimation and an MIP model for the optimization of the two areas at the same time. The objective is to maximize the total number of orders served with minimal impact on the average order delivery time. Extensive experiments using real-world data demonstrate that the proposed framework outperforms several benchmarks in current practice. 3. We focus on the optimization of order dispatching time for on-demand food delivery services by dynamically optimizing the time intervals for dispatching orders on such on-demand food delivery platforms. This study is motivated by a practical challenge encountered by a food delivery platform, wherein customer orders need to be allocated to couriers responsible for collecting food from designated centers and delivering it to customers within specified time windows. This setting poses a dynamic pickup and delivery problem where prompt delivery without delays is the critical objective. Specifically, we address this challenge by formulating the problem as a Markov decision process (MDP) and proposing a two-stage framework that integrates a multi-agent reinforcement learning (RL) approach for order dispatching and a heuristic method for courier routing. The multi-agent reinforcement algorithm determines the optimal timing for each order's entry into the matching pool, while the routing method incorporates orders into the couriers' delivery routes for pickups and deliveries. Extensive experiments were conducted, evaluating our approach using real-world data and a well-designed simulator. The results demonstrate the superior performance of our proposed framework compared to the currently practiced strategy. In summary, this thesis addresses new research problems arising from on-demand delivery, drawing upon new methods in AI. Experiments conducted with real-world urban delivery data demonstrate that our proposed data-driven optimization approaches can significantly enhance operational efficiency and reduce delivery costs. This thesis also opens various opportunities for future research, as discussed in the concluding chapter. Specifically, those emerging approaches that leverage machine learning and deep learning to develop optimization methods for vehicle routing problems from end to end for real-world urban delivery scenarios.