01659726

Component

Reliable sensor data, as required by many urban traffic management systems, can play a significant role in improving the efficiency, safety, and ultimately sustainability of a transportation system. One of the main challenges pertaining to sensor data is its quality assessment, as empirical research has shown inconsistency and missing data from a large portion of road sensors. Many previous studies focused on identifying ‘bad’ sensors whose data should be discarded and the recovery of traffic volume records is left to statistical imputation. This leaves two important questions yet to be addressed: how “bad” the malfunctioned sensor is, namely, how large the systematic error of a sensor is if it is not entirely failed; and how to recover traffic volume from a mix of accurate and erroneous information obtained by traffic sensors. The paper aims at addressing them by developing a moment matching based measurement error estimation method using spatial correlation of traffic counts from sensors in a roadway network. The authors contract the regular traffic network graph into a sensor network which highlights the correlation between individual sensor data and removes links and nodes that are irrelevant to sensor deployment. A nonlinear constrained mathematical program is formulated and solved to identify sensor systematic errors. In order to examine the effect of some key factors in the method performance, the authors provided numerical sensitivity analyses on two toy networks. A moderately large size example is used to demonstrate its capability of being employed in a more complicated network.

This paper was sponsored by TRB committee ABJ50 Standing Committee on Information Systems and Technology.

18-06744

English

19p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; References

Data collection; Data quality; Errors; Sensitivity analysis; Sensors; Traffic counts; Traffic volume

Data and Information Technology; Transportation (General)

Transportation Research Board Annual Meeting 2018 Paper #18-06744

TRIS, TRB, ATRI

Jan 8 2018 11:45AM