01627737

Component

Travel time is considered as one of the most important traffic variable both for systems’ users as well as operators, as can be identified by the industrial applications that have been presented in the recent years, and the accompanied research proposed on this field. In the current era of rich data availability, new opportunities and threats are emerging in modeling and predicting travel time. The scope of this paper aims to the introduction of reliability metrics in travel time predictions, additionally to the requirements for prediction robustness and accuracy. For achieving this, a novel dynamic and adaptive mechanism for predicting future travel time distributions (instead of single-values projections) is proposed, utilizing extensive and multi-source traffic datasets, able to stochastically and reliably map travel times in congested urban areas. The proposed framework is based on stochastic treatment of observed travel times that are conditioned by multiple traffic variables and collected from multiple sources (loop detectors and Bluetooth sensors), on a dynamic and adaptive control scheme. The methodological foundations of the proposed prediction mechanism rely on semi-parametric hazard-based modeling, suitably introduced in an adaptive calibration mode, able to project travel time distributions. Application results are presented in detail, using extensive datasets from the Cyprus Traffic Control Centre, exposing the notable benefits that can be obtained from the use of Big Data stochastic analysis in travel time predictions, especially in realistic circumstances.

This paper was sponsored by TRB committee ABJ30 Standing Committee on Urban Transportation Data and Information Systems.

01618707

17-03354

English

15p

2017

Transportation Research Board 96th Annual Meeting

Digital/other

Figures; References

Data fusion; Distributions (Statistics); Forecasting; Metrics (Quantitative assessment); Stochastic programming; Travel time

Data and Information Technology; Highways

Transportation Research Board Annual Meeting 2017 Paper #17-03354

TRIS, TRB, ATRI

Dec 8 2016 11:15AM