Trajectory grouping structure

K.A. Buchin, M.E. Buchin, M.J. van Kreveld, B. Speckmann, F. Staals

Research output: Contribution to journalArticleAcademicpeer-review

110 Downloads (Pure)

Abstract

The collective motion of a set of moving entities like people, birds, or other animals, is characterized by groups arising, merging, splitting, and ending. Given the trajectories of these entities, we define and model a structure that captures all of such changes using the Reeb graph, a concept from topology. The trajectory grouping structure has three natural parameters that allow more global views of the data in group size, group duration, and entity inter-distance. We prove complexity bounds on the maximum number of maximal groups that can be present, and give algorithms to compute the grouping structure efficiently. We also study how the trajectory grouping structure can be made robust, that is, how brief interruptions of groups can be disregarded in the global structure, adding a notion of persistence to the structure. Furthermore, we showcase the results of experiments using data generated by the NetLogo flocking model and from the Starkey project. The Starkey data describe the movement of elk, deer, and cattle. Although there is no ground truth for the grouping structure in this data, the experiments show that the trajectory grouping structure is plausible and has the desired effects when changing the essential parameters. Our research provides the first complete study of trajectory group evolvement, including combinatorial,
algorithmic, and experimental results.
Original languageEnglish
Pages (from-to)75-98
JournalJournal of Computational Geometry
Volume6
Issue number1
DOIs
Publication statusPublished - 2015

Fingerprint

Dive into the research topics of 'Trajectory grouping structure'. Together they form a unique fingerprint.

Cite this