Koopman mode decomposition for short-term traffic prediction


METADATA ONLY

Date

2025-05

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

Traffic flow data exhibits temporal and spatial correlations and a dynamic sequential structure; therby making short-term prediction challenging. This study explores the application of the Koopman operator for traffic prediction, which transforms a nonlinear system into a linear one in an infinite-dimensional space. The data-driven nature of Koopman mode decomposition (KMD) enables it to effectively capture these spatio-temporal correlations, making it well- suited for traffic prediction. Specifically, we incorporate known spatio-temporal inter-dependencies in traffic flow to develop a physics-informed modeling pipeline. A comparison between KMD and its linear counterpart, dynamic mode decomposition (DMD), demonstrates that KMD yields more accurate predictions and handles a wider range of traffic scenarios. Future research focuses on improving the robustness of KMD by addressing challenges related to missing and noisy data, further enhancing its applicability in real-world traffic prediction.

Permanent link

Publication status

unpublished

External links

Editor

Book title

Journal / series

Volume

Pages / Article No.

Publisher

STRC

Event

25th Swiss Transport Research Conference (STRC 2025)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Organisational unit

08686 - Gruppe Strassenverkehrstechnik check_circle

Notes

Conference lecture held on May 15, 2025

Funding

Related publications and datasets