Abstract
Freight vehicle tours and tour-chains are essential elements of state-the-art agent-based urban freight simulations as well as key units to analyse freight vehicle demand. GPS traces are typically used to extract vehicle tours and tour-chains and became available in a large scale to, for example, fleet management firms. While methods to process this data with the objective of analysing and modelling tour-based freight vehicle operations have been proposed, they were not fully explored with regard to the implication of underlying assumptions. In this context, we test different algorithms of stop-to-tour assignment, tour-type and tour-chain identification, aiming to expose their implications. Specifically, we compare the traditional stop-to-tour assignment algorithm using the location of a “base” as the start/end point of tours, against other algorithms using stop activities or payload capacity usage. Furthermore, we explore high-resolution tour-type/chain identification algorithms, considering stop types and recurrence of visits. For tour-chain identification, we explore two algorithms: one defines the day-level tour-chain-type based on the predominant tour-type identified for the period of 1 day and another defines the tour-chain-type based on the average number of stops per tour by stop type. For a demonstration purpose, we apply the methods to data from a large-scale GPS-based survey conducted during 2017–2019 in Singapore. We compare the algorithms in an assessment of freight vehicle operations day-to-day pattern homogeneity. Our analysis demonstrates that the predictions of tours, tourtypes, and tour-chain-types are highly dependent on the assumptions used, underlining the importance of carefully selecting and disclosing the methods for data processing. Finally, the exploration of day-to-day pattern homogeneity reveals operational differences across vehicle types and industries.
An Investigation of Parallel Road Map Inference from Big GPS Traces Data
