Preference-Aware Delivery Planning for Last-Mile Logistics

Qian Shao,Shih-Fen Cheng
2023-03-08
Abstract:Optimizing delivery routes for last-mile logistics service is challenging and has attracted the attention of many researchers. These problems are usually modeled and solved as variants of vehicle routing problems (VRPs) with challenging real-world constraints (e.g., time windows, precedence). However, despite many decades of solid research on solving these VRP instances, we still see significant gaps between optimized routes and the routes that are actually preferred by the practitioners. Most of these gaps are due to the difference between what's being optimized, and what the practitioners actually care about, which is hard to be defined exactly in many instances. In this paper, we propose a novel hierarchical route optimizer with learnable parameters that combines the strength of both the optimization and machine learning approaches. Our hierarchical router first solves a zone-level Traveling Salesman Problem with learnable weights on various zone-level features; with the zone visit sequence fixed, we then solve the stop-level vehicle routing problem as a Shortest Hamiltonian Path problem. The Bayesian optimization approach is then introduced to allow us to adjust the weights to be assigned to different zone features used in solving the zone-level Traveling Salesman Problem. By using a real-world delivery dataset provided by the Amazon Last Mile Routing Research Challenge, we demonstrate the importance of having both the optimization and the machine learning components. We also demonstrate how we can use route-related features to identify instances that we might have difficulty with. This paves ways to further research on how we can tackle these difficult instances.
Artificial Intelligence,Machine Learning,Multiagent Systems
What problem does this paper attempt to address?
The paper attempts to address the problem of optimizing delivery routes in last-mile logistics services. Despite decades of research, existing Vehicle Routing Problem (VRP) solutions still exhibit significant gaps in practical applications, primarily due to the differences between optimization objectives and the concerns of actual operators. Specifically, while traditional methods typically optimize metrics such as travel time, distance, or cost, experienced delivery drivers, based on their knowledge of the service area and customers, may plan routes according to various factors that are difficult to formalize and quantify. As a result, drivers often deviate from planned routes, which can impact overall system metrics such as fuel consumption, delivery time, and order loading operations. To address this issue, the authors propose a novel hierarchical routing optimizer that combines the strengths of optimization methods and machine learning approaches. This optimizer first solves a Traveling Salesman Problem (TSP) with learnable weights at the regional level, and then, after determining the sequence of regional visits, solves the site-level VRP within each region. By introducing Bayesian optimization methods, the weights assigned to different regional features can be adjusted to better reflect the preferences of actual operators. Using a real-world delivery dataset provided by the Amazon Last Mile Routing Research Challenge, the authors demonstrate the importance of combining optimization and machine learning components and explore how route-related features can be used to identify potentially difficult instances.