Traffic Signal Control Algorithm Based On Traffic Flow Prediction

With the rapid development of economy and the sharp increase of motor vehicle ownership,urban traffic congestion is becoming increasingly serious,especially at intersections.Traditional timing traffic lights can not be very good at some intersections where the traffic flow varies greatly with the change of time.How to effectively relieve the traffic pressure at intersections and reduce traffic congestion has become the primary problem.The short-time traffic flow prediction has the characteristics of real-time,nonlinear and periodicity,so the combination of traffic flow prediction and traffic signal dynamic control provides a solution to relieve traffic pressure.Firstly,by comparing different models through experiments,the long-term and short-term memory neural network(LSTM)is finally used to predict traffic flow.In order to avoid the phenomenon of overfitting and gradient explosion,a combined prediction model(LSTM-KF)combining LSTM and Kalman filter(KF)is proposed in this paper.The dropout layer is introduced into LSTM,and Adam is selected as the optimizer.The optimal experimental parameters are selected by repeated comparison in the experiment.Comparing the experimental results with the same index,the results show that the combined prediction model proposed in this paper is superior to the single prediction model.Secondly,the combination of traffic flow prediction and traffic signal dynamic control is realized.Based on the prediction data,the deep reinforcement learning algorithm DQN model was used for dynamic control of traffic signal duration,and the experimental parameters of the DQN model were determined through analysis.Traffic simulation software(SUMO)is used to compare the DQN traffic signal dynamic control scheme and fixed timing scheme used in this paper.The simulation results show that the DQN model used in this paper is superior to the fixed timing of traffic signal dynamic timing.Finally,a traffic signal control system based on the prediction algorithm and traffic flow control algorithm is built.The system realizes three modules respectively: real traffic flow data & forecast traffic flow data management application,traffic flow forecast application time control application and traffic signal time visualization management,providing convenience for users.In this paper,the design and implementation of traffic signal control algorithm based on traffic flow prediction is proposed to relieve the traffic pressure at intersections and make the traffic signal time control more elastic.At the same time,it provides enlightenment for other research on relieving traffic pressure,especially on reducing traffic pressure at intersections.