The motive for Amsterdam to start this project was the omission of the van segments “construction” and “service” in the old module. These are vans that are driven by, for example, construction workers to a building site or mechanics to a repair job. Besides adding these two van segments, improved modules were asked for modelling freight trucks and for van trips for goods and parcel transportation.
The developed module consists of three sub modules: the freight module (VR), the service / construction van module (BE) and the parcel delivery module (PK).
The freight module calculates the number of departing and arriving freight truck trips for each zone of the VMA module using a regression equation that considers (1) the number of jobs by sector, (2) the number of inhabitants, (3) the surface of distribution centers and parcel fulfillment centers and (4) whether a zone is port area or an airport (in Amsterdam the airport Schiphol and port area Westpoort play a large role in goods transportation). Based on the number of departures and arrivals in each zone, a gravity model calculates the number of trips between the zones.
The module for service and construction vans works in a similar fashion as the freight module. The regression equations are estimated separately for the service segment and the construction segment. For example, the number of inhabitants and the number jobs in the retail segment are more important in explaining the generation of service trips than the generation of construction trips.
Finally, there is the module for parcel deliveries. In contrast tot the previous two modules, this module works at the level of individual round tours departing from fulfillment centers, rather than aggregate numbers of trips between zones. Based on statistics of the Netherlands Authority for Consumers and Markets, factors were deduced that allow for calculation of the number of B2B and B2C parcels in each zone and assignment of these parcels to the six largest parcel couriers of the Netherlands. Next, the parcels are assigned to one of the fulfillment centers of the respective couriers, after which round tours are formed using optimization heuristics.
Already before calibration on traffic counts, the new module shows good results. We see, for example, that more trips are produced in the Westpoort port area, where the old module showed a clear underestimation, and the match on traffic counts (before calibration) is improved. The first follow-up step will be the calibration of the model on traffic counts.