Abstract

Capacity at signalized intersections is a basic parameter in urban transport networks. The capacity of a signalized intersection is a function of existing geometric, control, weather, and other conditions. Estimation of capacity at signalized intersections is one of the most important topics in traffic Engineering. If the capacity can directly be measured, the delay at signalized intersections and thus the traffic performance and quality of service can be calculated according to the functional relationship between delay and capacity. Unfortunately, under real world traffic conditions, the capacity cannot easily be measured directly for an existing intersection, especially under unsaturated flow conditions where the flow volume is lower than the capacity. This paper presents a model for estimating the capacity of an existing signalized intersection under unsaturated flow condition based on the cycle overflow probability which can be directly measured by loop detectors at stop lines. The cycle overflow probability is just the proportion of cycles with occupied detector at end of green time phases. Also the flow volume can directly be measured by loop detectors at stop lines. According to the queuing theory, the cycle overflow probability is a function of the degree of saturation, i.e. a function of flow volume and capacity. Thus, by measuring the cycle overflow probability and the flow volume, the capacity can be estimated according to the functional relationship. The proposed model is calibrated to data obtained from loop detectors at different signalized intersections under unsaturated conditions. For validation the model, the real capacities at the same intersections are measured for saturated cycles where the capacity can be considers as the measured volume. The proposed model delivers a useful tool for estimating capacity at signalized intersections under unsaturated conditions. Using the proposed model, the capacity and thus the traffic quality of service at existing signalized intersections can directly be estimated using data from loop detectors at stop lines. The model is theoretically reasonable and easily to use for practitioners. The results of the calibration and validation are very promising.