Estimation of time-dependent intersection turning proportions for urban signal controls

Abstract

To implement an adequate adaptive traffic signal control strategy, one of the critical components is the reliable estimates of intersection turning proportions. However, in practice link traffic counts are not sufficient to provide desirable turning proportion estimates at an intersection due to the underdetermined problem and it needs to further consider other auxiliary data sources to improve the estimation accuracy. Based on a traffic flow model, a nonlinear least squares estimate model is proposed to integrate link flow information from vehicle detectors (VDs) and partial turning flow information from video sensors to solve the turning proportion estimation problem. Using the simulation tool, DYNASMART-P, as a test platform, the test results indicate that heterogeneous data resources are better than single source, and this framework has potentials to feedback to an on/off-line adaptive traffic signal control.