Optimizing actuated traffic signal control using license plate recognition data: Methods for modeling and algorithm development

Abstract

Primary control parameters (e.g., minimum and maximum green times) at actuated traffic signal controllers are traditionally determined based on the limited historical or sample traffic data, along with engineering experience, which actually cannot well respond to the real-time traffic demands in operation. Meanwhile, License Plate Recognition (LPR) data detected by Electronic Police devices have become more and more available at urban signalized intersections in China. Therefore, this paper presents a novel idea in developing a set of innovative algorithms to dynamically determine the control parameters against varying traffic conditions measured by the LPR data. The designated computational loop is developed to determine from the LPR data: (1) traffic parameters (e.g., traffic arrivals of the target intersection in the future unit time) dynamically at small time intervals; (2) traffic arrival states and initial states of all the approaches at the subject intersection using the estimated traffic parameters from previous time interval, and then predict the traffic states for the next time interval; (3) traffic signal optimization alternative to maximize the total throughput while minimizing the overall control delay. At the same time, the delay and queue length of the subject intersection are used as a performance feedback to the next cycle of calculation for continually optimizing the actuated control timing parameters. The simulation-based comparison of the improved actuated signal control scheme with the existing pre-timed control scheme indicates an obvious improvement by the developed algorithms based on the LPR data obtained at typical intersections in Jinan, Shandong Province, China.