Evaluating city logistics measures using a multi-agent model

TLDR

The study proposes a multi‑agent methodology to evaluate city logistics measures by modeling stakeholder behaviour in urban freight transport. The approach combines Q‑learning for adaptive behaviour with a VRP‑TW‑F vehicle routing and scheduling model, implemented on a simulated urban road network. The simulation shows that jointly banning trucks in polluted zones and eliminating motorway tolls yields significant environmental benefits and stakeholder‑acceptable outcomes.

Abstract

This paper presents a methodology for evaluating city logistics measures considering the behaviour of several stakeholders associated with urban freight transport using a multi-agent model. The model constructed consists of a learning model and a model for vehicle routing and scheduling problem with time window-forecasted (VRP-TW-F). We used a method of Q-learning, a technique of reinforcement learning, in constructing a learning model. We implemented the model on a test road network representing an urban area. The results indicate that implementing a truck ban directly to environmentally damaged areas and discounting motorway tolls entirely in the urban motorway network together has large environmental effects, and leads to an acceptable environment for all stakeholders.