Method for Real-time Delay Estimation at Signalized Intersection

The typical problems of emerging methods based on detector data for measuring delay at urban signalized intersections such as input hypothesis, model establishment and fault -tolerance performance were pointed out. A new method named "Probability sampling" which use two rows of detectors for dynamic survey of approach delay on signal cycle basis was put forward, including the process of cycle split, the arrival time of first car and the departure time of last car at exit detectors, the upper and lower boundary of time interval at entrance detectors, and matching method for vehicular arrival and departure. Finally, a simulation model on Wenyi-Jiaogong Intersection and an actual intersection in Shanghai (Cao'an-Qilianshan Intersection) were tested to verify the accuracy of the proposed method. The results show that Mean Absolute Error (MAE) of the measurement delay is 4.3 seconds under no detector error and 4.9 seconds under 5% detector error rate. The method is insensitive to the detector error, and can be used to provide the accurate, timely and reliable delay index for traffic control system.