Bottleneck identification locates problematic segments on a freeway corridor and meanwhile provides information about the cause and characteristicsof the congestion. It is a critical step in mitigating the urban congestion problem. Due to the wide availability of traffic surveillance data, researchers have been designing bottleneck identification algorithms based onarchived traffic flow data. Those algorithms include rule-based, contour-map-based and simulation-based methods. However, these existing methods require traffic data with high accuracy and consistency, which may not always be the case in reality. In this paper, a new bottleneck identification method based on coordinate transformation on fundamental diagram is proposed. The algorithm is designed for fix-location detector data and can tolerate noise and inconsistency. Three loop detector datasets were collected atthe city of Madison and the city of Milwaukee, WI, USA. The three datasets have different levels of data quality so that the effectiveness and robustness of the proposed algorithm can be tested. Meanwhile, a novel evaluation strategy for bottleneck identification in the absence of ground truth data was first introduced in this paper. Using this strategy, the proposedalgorithm is compared with Chen’s method. The evaluation results indicatesuperior effectiveness and robustness of the proposed algorithm comparingto earlier methods.
Abstract