A data-driven framework for characterising road freight transport using GPS data

Artikel i vetenskaplig tidskrift, 2026

Granular insights into road freight transport are essential for evidence-based policy-making, yet a persistent gap remains between regulatory frameworks and policy objectives due to a lack of disaggregated data. While GPS data offers a low-resource pathway to detailed truck movement analysis, existing literature focuses heavily on stop identification, leaving systematic, national-scale characterisations of road freight transportation systems relatively scarce. This study addresses this gap by developing a data-driven framework that utilises GPS data to comprehensively characterise the road freight transport system across four dimensions: truck flows, network, trips, and activities. Because raw GPS data lacks event descriptors, we develop a multi-step processing approach using machine learning to identify truck stops, which then serve as the foundation for the characterisation. We utilise a decision tree model with thresholds calibrated via sequential unsupervised clustering, bridging natural data patterns with domain expertise. The model's internal consistency is assessed through a cross-model agreement audit with a Random Forest classifier. We demonstrate the framework's utility through a nationwide case study in Belgium using tolling data from heavy goods vehicles. To assess the results’ external plausibility, the GPS-inferred metrics are compared against national aggregated data. This systematic characterisation of a country's road freight transport system provides a scalable foundation for targeted policy-making, enabling a deeper understanding of its challenges and the evaluation of measures to improve transport sustainability.