Last: Mile route optimization with machine learning

Chan, Tze Keet (2024) Last: Mile route optimization with machine learning. Final Year Project, UTAR.

Preview

PDF Download (2993Kb) | Preview

Abstract

Ever since COVID-19 pandemic, online shopping had been skyrocketed. To handle the enormous volume of deliveries, last-mile delivery route planning and optimization had become more significant than ever for logistics services. Last-mile logistics are referring to the final stage of the delivery process, where goods are transported from a distribution hub to the end destination, typically a residential or commercial address. Last-mile logistics had always been the costliest part in the overall supply chain. Numerous last-mile route optimization models/frameworks are proposed and been practiced in logistics services, to reduce operation costs while attempt to fulfill customers’ satisfaction. However, existing pure optimization frameworks often overlooked that in real-world practices, the prescribed routes may be not followed by delivery drivers, as they may prioritize personal knowledges and experiences. Deviation of prescribed delivery routes by delivery drivers may be due to various underlying reasons, including but not limited to traffics conditions, and customers’ preferences. In this project, we proposed a Simple R-NN model to uncover the underlying relationship/pattern between customers’ acceptable delivery time windows and deviations of prescribed delivery routes by drivers. The proposed model, Simple R-NN model aims to predicts possible delivery routes by drivers, then output an optimized delivery route that seems acceptable for the drivers to actual adapts in actual delivery operation.

Actions (login required)