Traffic State Estimation in Congestion to Extend Applicability of DFOS

TL;DR

This paper introduces a data assimilation-based method to impute missing vehicle velocity data in DFOS, enabling effective traffic state estimation during severe congestion.

Contribution

It presents a novel missing data imputation technique for DFOS, improving its reliability in low-visibility traffic conditions.

Findings

Imputation method maintains low MAE in velocity estimates.

Validated on Japanese expressways with successful results.

Enhances DFOS applicability in severe congestion scenarios.

Abstract

This paper presents a traffic state estimation (TSE) method in congestion for distributed fiber-optic sensing (DFOS). DFOS detects vehicle driving vibrations along the optical fiber and obtains their trajectories in the spatiotemporal plane. From these trajectories, DFOS provides mean velocities for real-time spatially continuous traffic monitoring without dead zones. However, when vehicle vibration intensities are insufficiently low due to slow speed, trajectories cannot be obtained, leading to missing values in mean velocity data. It restricts DFOS applicability in severe congestion. Therefore, this paper proposes a missing value imputation method based on data assimilation. Our proposed method is validated on two expressways in Japan with the reference data. The results show that the mean absolute error (MAE) of the imputed mean velocities to the reference increases only by 1.5 km/h as compared with the MAE of non-missing values. This study enhances the wide-range applicability of DFOS in practical cases.