Light goods vehicles are an important part of London traffic. With changes to delivery demand and traffic patterns more broadly, they often have a big impact on cities. A group of researchers partnered with industrial organizations specializing in last-mile parcel delivery, thereby gaining access to data which allowed researchers to construct an agent-based model of the last-mile delivery process.
In their paper, the researchers expand upon the existing model to incorporate parking behavior, an important factor of delivery driving which is often overlooked in the literature. The tool they present can be used to explore different policy and infrastructure interventions.
Walking or driving?
Perhaps most unexpectedly, delivery drivers were spending on average more than half of the time on their rounds with the vehicles parked. They were also walking 7.9 km per round on average, compared to driving 11.9km within the target area. A minimum of 80% of parking stops took place in street parking situations rather than any kind of specialized loading/unloading area. These results suggested that the drivers were going about the delivery process rather differently than expected. Parking and its availability were dynamic and influential factors in the business of delivery.
Delivery performance and driver behavior
This work has presented a technique for extending an existing agent-based framework to include a new and significant behavior, namely parking. In doing so, we have incorporated aspects of the physical infrastructure into consideration, as well as driver knowledge and behavior. The inclusion of these important aspects of delivery performance allows to more meaningfully delve into how interventions in parking regulations, driver training, or infrastructure redevelopment might influence the system overall.
Future work might expand upon this by incorporating more complex driver parking behaviors. In particular, the researchers note that the heuristic by which drivers make their deliveries is extremely naive, and may well not track with the actions of the individual drivers. By comparing the generated, synthetic outputs of the driver choice process with the experiences of real drivers, it may be possible to better frame conversations with subject matter experts and crystallize aspects of driver behavior for further study.
The model presented makes the outcome of different policies and investments a topic which urban planners can more meaningfully compare and discuss. By helping non-specialists explore different trade-offs between proposed interventions, we can help both urban planners and industry partners understand both how things are now and how they may develop in the future. These changes may have implications on the health of the city, both in terms of the well-being of its citizens and the viability of its economic engine.