Abstract

Driving cycles are required for a variety of applications including longitudinal dynamics simulations. For the generation of representative driving cycles, a driving data analysis is indispensable. This paper proposes a method to efficiently segmenting data and subsequently identifying typical trip sections. A first cluster analysis is performed on individual data points using the kmeans++ algorithm. Based on the results, the consecutive data points are segmented into microsegments. Subsequently, these microsegments are being clustered in a second cluster analysis. The results obtained reveal patterns of cluster formations that are similar to those observed in the cluster analysis of individual data points.

Another segmentation, based on the minimum duration of standstill times between two driving sections, enables the identification of typical trips of longer durations. This is achieved by taking the proportions of the microsegments assigned to the same cluster as input variables for the third cluster analysis. Thereby, groups of similar trips can be identified with the typical distribution of microsegment proportions.

Thus, the developed method yields representative trip sections for a driving dataset and thereby forms a basis for generating representative driving cycles both in the research and in the development of simulationbased technologies.