Correction of Bias Errors in Detected Traffic Volume on a Virtual Urban Expressway Network

Abstract

There are cases where detected traffic volume includes bias errors of ten percent or more due to the combined factors of changes to vehicle detectors and lane-changing vehicles. These bias errors can considerably lower the precision in setting important parameters for traffic simulations because of the failure to account for the law of conservation of vehicles. Therefore, detection and correction of the bias errors are essential. In this study, bias correction coefficients for free flow as well as for congested flow were estimated at every cross-section where traffic detectors were installed on a virtual urban expressway network with a circular section and four leg junctions. In the estimation, a simulated traffic volume that included random and bias errors was used. The findings showed a possibility that estimation errors of bias correction coefficients could be restricted to a certain level if the estimated error variance as an output of the Kalman filter was used as one criterion.