01552830

Component

Real-time traffic condition is a critical input for modern intelligent transportation systems (ITS). However, current real-time traffic state estimators are all link-based with the assumption that the traffic condition is homogeneous across multiple lanes. This assumption helps in designing the estimators but is insufficient for many occasions, e.g., toll lanes. On the other hand, the data-driven approach has the potential to be used for lane-based estimation but incurs huge computational cost, making it hard to be implemented on-line. In addition, although many traffic sensing technologies are available, most of the estimators utilize only one type of measurements because of the difficulties in combining heterogeneous data. This paper proposes a multi-sensor data fusion framework for real-time lane-based traffic state estimation. A bi-level architecture is adopted to combine a model-based approach and a data-driven approach to keep the computation cost low while enabling the lane-based estimation. A spatial-temporal smoothing filter is developed which can conveniently incorporate heterogeneous measurements. Simulation-based analysis shows that our approach is effective.

This paper was sponsored by TRB committee ABJ35 Highway Traffic Monitoring.

01550057

15-0186

English

2015

Transportation Research Board 94th Annual Meeting

Digital/other

Figures; References; Tables

Algorithms; Data fusion; Heterogeneity; Image processing; Remote sensing; Traffic estimation; Traffic lanes

Data and Information Technology; Highways; I72: Traffic and Transport Planning

Transportation Research Board Annual Meeting 2015 Paper #15-0186

TRIS, TRB, ATRI

Dec 30 2014 12:13PM