Novel Evaluation Method for Signalized Intersections

Danya Yao,Yi Fu,Wei Guo,Yi Zhang,Renjie Teng

2006-01-01

Abstract:Evaluating performance of signalized intersections is very important in traffic congestion management. Some typical methods, such as Highway Capacity Manual's average vehicle delay based Level of Service estimation method, have been proposed, and these methods use vehicle delay to index the commuters' time waiting and discomfort feeling directly. However, vehicle delay is often estimated approximately by various mathematical models and uses a few detection traffic data. After traffic detecting systems have been installed, it is easy to acquire and store traffic data. But few previous evaluation methods have adequately made use of the massive data, especially the historical traffic data. In this paper, a novel performance evaluation method based on data mining techniques for signalized intersections is proposed. This method represents the operating condition of signalized intersections based on a so-called pseudo traffic flow-occupancy model. Expectation Maximization (EM) algorithm is adopted to extract features from massive data. Then Congestion Changing Index (CCI) is defined to indicate the change of congestion level quantificationally. Experiments carried on typical signalized intersections show that this proposed model method and feature extracting algorithm is feasible, the CCI result is effective especially when it is carried out on evaluating the congestion level changing after control scheme adjusting. All of these experiments data are from loop detectors of Beijing's urban traffic control system.