Agent-based model of last-mile parcel deliveries and travel demand incorporating online shopping behavior

Abstract

In this paper, we present an extension of the agent-based travel demand model mobiTopp with a last-mile parcel delivery module called logiTopp, in which online shopping choice is modeled explicitly. Online shopping behavior is modeled using logistic and Poisson regression models, which consider both the socio-demographic characteristics of the customer and aspects of their travel behavior. As mobiTopp is a framework that simulates travel demand over one week, we are able to capture interactions between travel behavior and online shopping that do not become apparent in single-day simulations.

The results show that the integrated choice model reflects the findings presented in the literature in that male, affluent, young professionals are most likely to (frequently) order parcels online compared to other groups of the population. Application of the agent-based model to a city in Germany shows that socio-demographic and behavioral characteristics are considered realistically within the simulation.

The model presented here is a suitable simulation tool for alternative urban last-mile delivery solutions, and the open-source and modular framework allows for transfer to other regions as the underlying choice models are consistent with literature from other spatial contexts.

The findings are of interest to transportation planners and policymakers as they contribute to the understanding of how increased e-commerce demand influences the transportation system and solutions to mitigate adverse effects.