Abstract:Against the backdrop of the COVID-19 pandemic since 2019, characterized by the implementation of stringent quarantine restrictions in public cities and a general fear of virus transmission, there has been a significant increase in global demand for e-commerce services [8–11], particularly doorstep delivery of goods. The practice of delivering goods directly from stores to doorsteps, even within city neighborhoods, has gained widespread adoption [9; 10], especially in developed countries. Consequently, the market for courier and postal services has substantially grown [12; 13]. The active use of courier services and the relaxation of quarantine measures, leading to a return to pre-pandemic levels of public and personal transport usage, have overtime increased the strain on urban road infrastructure. While the purchasing power and attitudes of consumers toward large expenses have decreased due to rising global oil prices and, accordingly, increased fuel costs [14], the demand for courier and postal services as a whole has remained steady. This is attributed to the growing trend of demand for the delivery of essential food and goods amid restrictions, as was seen during the pandemic [9]. Additionally, competition in this market is continually on the rise [15]. These factors pose a series of challenges for service providers, one of which is improving the methods and algorithms for calculating delivery routes within cities, as many of these providers also cover last-mile delivery. This is primarily necessary to minimize ancillary costs (fuel, vehicle maintenance, etc.) and, consequently, to offer a better proposition to users. In most cases, the calculation of delivery routes resembles the classic "Traveling Salesman Problem." Most existing algorithms for route calculation rely on static data. Today, such calculations for city deliveries can quickly become outdated due to various factors such as traffic jams caused by various circumstances, adverse weather conditions, and more. The main challenge with classic algorithms is the need for real-time route adjustments, which require constant recalculations. Although classic algorithms efficiently compute optimal routes, they are quite slow when it comes to adjusting routes in real-time. One solution to address the issue of efficient route calculation is the application of Artificial Intelligence (AI) and Machine Learning (ML) techniques and incorporating ML elements into the adjustment of calculation results through conventional algorithms or fully relying on AI for efficient route planning. This research is conducted in the context of C2C (Customer to Customer) delivery of small-sized cargo. The possibility of improving the calculation of efficient routes between pick-up and drop-off points on a single route is explored. This scheme involves parcel collection and delivery as an ancillary process for the courier, such as in scenarios where an individual wishes and is able to deliver some parcels along their route to compensate for their personal transportation expenses. Therefore, this study examines the possibilities and practices of enhancing the calculation of such routes using AI and ML based on previous related research.

Can language models be used for real-world urban-delivery route optimization?

Can Large Language Models Solve Robot Routing?

Understanding Drivers' Route Choice Behaviours in the Urban Network with Machine Learning Models

Research on Travel Route Planning Optimization based on Large Language Model

Joint Estimation and Prediction of City-wide Delivery Demand: A Large Language Model Empowered Graph-based Learning Approach

Towards Next-Generation Urban Decision Support Systems through AI-Powered Construction of Scientific Ontology using Large Language Models -- A Case in Optimizing Intermodal Freight Transportation

AI-Driven Day-to-Day Route Choice

Receive, Reason, and React: Drive as You Say, With Large Language Models in Autonomous Vehicles

Preference-Aware Delivery Planning for Last-Mile Logistics

An integrated route planning approach for driverless vehicle delivery system

Probability estimation and structured output prediction for learning preferences in last mile delivery

Receive, Reason, and React: Drive as You Say with Large Language Models in Autonomous Vehicles

Optimising Stochastic Routing for Taxi Fleets with Model Enhanced Reinforcement Learning

A novel multi-objective optimization model for the vehicle routing problem with drone delivery and dynamic flight endurance

A deep reinforcement learning with dynamic spatio-temporal graph model for solving urban logistics delivery planning problems

SurrealDriver: Designing Generative Driver Agent Simulation Framework in Urban Contexts based on Large Language Model

Learning from Drivers to Tackle the Amazon Last Mile Routing Research Challenge

Analysis of practices in applying artificial intelligence and machine learning to improve urban route planning results

Large Language Models for Supply Chain Optimization

Distance VS. Coordinate: Distance Based Embedding Improves Model Generalization for Routing Problems

Data-driven optimization for last-mile delivery