Microscopic Simulation of Urban Last-Mile Delivery Impacts on Traffic Performance: A Case Study of Dammam, Saudi Arabia

Introduction: The surge of last-mile delivery services has implications for the composition and performance of urban road networks. This study explores the transportation impacts of the growing demand for deliveries using a geospatially grounded microscopic simulation of the city of Dammam, Saudi Arabia. 

Methods: A digital representation of the road network of Dammam was obtained from OpenStreetMap and processed into a Simulation of Urban MObility (SUMO) network. The network was then used to generate five replications of each of four delivery-demand scenarios with an added number of 200, 400, 600, and 800 delivery vehicles beyond those present in the original network. Each scenario was replicated five times, resulting in twenty microscopic traffic simulations. Travel time, waiting time, time loss, mean speed, and delivery completion rate were extracted as performance indicators (PIS). Additionally, the impact on background passenger traffic was estimated. Differences between the demand scenarios at the replication level were evaluated using the Friedman test. 

Results: An increase in delivery demand by 200–800 vehicles resulted in a 7.33% rise in the mean delivery travel time, 25.95% increase in waiting time, and 17.47% rise in time loss, whereas the mean delivery speed was lower by 2.71%. Furthermore, the delivery completion rate decreased by 3.06%, from 99.05% to 95.99%. Delivery travel time (p = 0.0189), time loss (p = 0.0263), and speed (p = 0.0050) were found to differ significantly between the demand scenarios. Concurrently, background traffic experienced a 3.61% rise in passenger travel time, 15.43% increase in waiting time, and 10.77% rise in time loss. These differences were also found to be significant across the demand scenarios. 

Conclusions: The current study found that intensifying last-mile delivery activities have non-negligible transportation consequences at the network level, with travel-time-related PIS being particularly sensitive to increased deliveries. Future research will calibrate the framework using observed Dammam traffic data and integrate reinforcement learning for dynamic delivery routing and vehicle assignment under varying traffic conditions.