Research On Urban Traffic Signal Control System Based On Traffic Flow Prediction

With the development of economy and the progress of society,urban traffic congestion is becoming more and more global.How to improve travel efficiency and alleviate traffic congestion has long been a key problem in intelligent transportation systems.Traffic signal control is the basic tool of urban traffic management,and traffic flow prediction provides the basis of traffic signal controlling.However,the current research on traffic signal control and traffic flow prediction is separated from each other.Neither pure optimization nor pure prediction can achieve ideal urban traffic control.To this end,the following work has been done in this paper:Firstly,this paper proposes the framework of an urban traffic signal control system based on traffic flow prediction,which combines traffic flow prediction and traffic signal controlling.The system includes data center,prediction module and optimization module.In the data center,the road traffic data in different formats can be converted to each other,and the traffic data that is difficult to be obtained by observation and statistics,such as OD traffic volume,are inferred.Secondly,the channel between prediction module and optimization module is established.The prediction model is trained by BP neural network with historical traffic data.The real-time traffic data becomes the prediction data used by the optimization module after the prediction model.The optimization module then optimizes the signal timing scheme which is most suitable for the prediction data and applies it to the road signal in the future.Finally,two adaptive control strategies,precise control optimization strategy and evolutionary control optimization strategy,are designed for different size of road networks.In order to validate the proposed system and optimization strategy,experiments are performed on the real-world traffic data provided by the Aliyun Tianchi platform.The experimental results show that the proposed optimization strategies and the prediction module of the system have obvious effect on improving the capacity of the intersection,and the system running time meets the timeliness requirements of the prediction.