Mechanistic Identification of Freight Activity Stops from Global Positioning System Data

Abstract

The identification of freight pick-ups and deliveries, referred to as “freight activity” in this paper, is crucial to characterizing freight operations and assessing the performance of freight transportation systems. However, identifying freight activity stops from global positioning system (GPS) data is challenging, particularly in urban freight where congested traffic is common. This paper presents a mechanistic—because it is based on the physics of driving patterns—procedure to identify freight activity stops from raw GPS data. The procedure was implemented to identify stops in three distinct case studies that present a wide range of traffic conditions: Barranquilla, Colombia; Dhaka, Bangladesh; and New York City, United States. The results show that the procedure achieves an average accuracy of above 98.6% when identifying freight activity stops. The results of the proposed procedure were compared with results from support vector machines, random forest, and k nearest neighbors. The mechanistic procedure outperformed these methods in correctly classifying freight activity using second-by-second GPS data.