A Multi-Objective Traffic Signal Control Model for Intersection Based on B-P Neural Networks

Real-time traffic counts between entry and exit of intersections are important input data for traffic signal control systems, and also difficult to obtain. In on-line systems, it is impossible to detect such data directly through a traffic surveillance system and also unrealistic to gather them from conducting surveys. This paper first reviewed the evolution process of the traffic signal control system and analyzed characteristics of real-time traffic counts at intersections. Then a back-propagation neural network (BPN) model was proposed to estimate and forecast the dynamic turning movements, and an algorithm was designed to solve the model. Furthermore, based on the estimated turning movements, a multi-objective model was put forward to achieve the real-time signal control parameters and the evaluation indices. Simulation experiments and case study were implemented and the results were compared with classical Webster method. The reported examples have clearly indicated that the models and algorithms are rather efficient and accurate, and can realize the real-time signal control for intersections.